Transcript
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
Issue
1.0
HUAWEI TECHNOLOGIES CO., LTD.
Copyright © Huawei Technologies Co., Ltd. 2016. All rights reserved. No part of this document may be reproduced or transmitted in any form or by any means without prior written consent of Huawei Technologies Co., Ltd.
Trademarks and Permissions and other Huawei trademarks are trademarks of Huawei Technologies Co., Ltd. All other trademarks and trade names mentioned in this document are the property of their respective holders.
Notice The purchased products, services and feat ures are stipulated by the contract made bet wee n Huawei and the customer. All or part of the products, services and features described in this document may not be within the purchas e scope or the usage scope. Unless otherwis e specified in the contract, all statements, information, and recommendations in this document are provided "AS IS" without warranties, guarantees or representations of any kind, either express or implied. The information in this document is subject to change without notice. E very effort has been made in the preparation of this document to ensure accuracy of the contents, but all statements, information, and recommendations in this document do not constitute a warranty of any kind, express or implied.
Huawei Technologies Co., Ltd. Address:
Huawei Industrial Base Bantian, Longgang Shenzhen 518129 People's Republic of China
Website:
http://www.huawei.com
Email:
[email protected]
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
i
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
About This Document
About This Document Intended Audience This document is intended for sale engineers, channel distributors, service managers, and customers who want to know how to deploy Huawei SAP HANA appliance to achieve real-time data processing ability.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
ii
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
Contents
Contents About This Document.....................................................................................................................ii 1 Overview of SAP HANA ............................................................................................................. 1 1.1 What Is SAP HANA ........................................................................................................................................................................ 1 1.2 Application Scenario of SAP HANA ........................................................................................................................................... 2 1.3 Systems Interworking of SAP HANA .......................................................................................................................................... 3 1.4 Logical Function Architecture of SAP HANA ........................................................................................................................... 4 1.5 Product Portfolio of SAP HANA .................................................................................................................................................. 5 1.6 Overview of SAP HA NA Appliance Cluster System ................................................................................................................ 5 1.6.1 Main Concepts of SAP HANA Appliance Cluster System ................................................................................................... 5 1.6.2 So ftware Deploy ment of SAP HANA Appliance Cluster System ....................................................................................... 7
2 Overview of Huawei SAP HANA Product .............................................................................. 1 2.1 Product Portfolio .............................................................................................................................................................................. 1 2.2 Product License ................................................................................................................................................................................ 2 2.3 Scope of this document ................................................................................................................................................................... 2
3 System Composition of Huawei FusionCube for SAP HANA Appliance ........................ 3 3.1 Overall Architecture ........................................................................................................................................................................ 3 3.2 Converged Devices for Co mputing, Storage and Network ...................................................................................................... 4 3.2.1 E9000.............................................................................................................................................................................................. 4 3.2.1.1 E9000 High-Density Subrack ................................................................................................................................................. 4 3.2.1.2 High-Density Blade of E9000 ................................................................................................................................................ 5 3.2.1.2.1 Ivybridge EX-based CH242 V3 Co mputing Blade .......................................................................................................... 5 3.2.1.2.2 CH222 Management Blade .................................................................................................................................................. 5 3.2.1.2.3 CH222 Storage Blade ............................................................................................................................................................ 6 3.2.1.2.4 CH223 Storage Blade ............................................................................................................................................................ 6 3.2.1.3 High-Performance Switching of E9000 ................................................................................................................................ 7 3.2.1.3.1 Ethernet Switch ...................................................................................................................................................................... 7 3.2.1.3.2 FC Switching .......................................................................................................................................................................... 8 3.2.1.3.3 IB Switching ........................................................................................................................................................................... 9 3.2.2 RH8100 V3 .................................................................................................................................................................................... 9 3.2.2.2 Product Features...................................................................................................................................................................... 10 3.2.2.3 Product Specifications............................................................................................................................................................ 12 3.3 Shared Storage Device .................................................................................................................................................................. 14
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
iii
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
Contents
3.3.1 Overview of NAS fro m Huawei .............................................................................................................................................. 14 3.3.1.1 Overview of p roduct............................................................................................................................................................... 14 3.3.1.2 Characteristics of product...................................................................................................................................................... 15 3.3.1.3 Specification of product......................................................................................................................................................... 16 3.3.2 NAS in FusionCube SAP HANA appliance .......................................................................................................................... 18 3.3.2.1 Main Co mpenonts................................................................................................................................................................... 18 3.3.2.2 Main Configurations .............................................................................................................................................................. 19 3.4 High-Performance Distributed Storage Sub-System ............................................................................................................... 19 3.4.1 High-Performance Snapshots ................................................................................................................................................... 20 3.4.2 High-Performance Lin ked Cloning ......................................................................................................................................... 21 3.4.3 Automat ic Thin Provisioning.................................................................................................................................................... 22 3.5 Characteristics of Huawei FusionCube for SAP HANA Appliance...................................................................................... 23
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance .................................. 24 4.1 Cluster Solution.............................................................................................................................................................................. 24 4.2 Data Persistence Solution ............................................................................................................................................................. 25 4.2.1 Working Princip le ....................................................................................................................................................................... 25 4.2.2 Media for Storing Persistent Data............................................................................................................................................ 26 4.2.3 Cache Accelerat ion Mechanism............................................................................................................................................... 27 4.3 Service Continuity Assurance...................................................................................................................................................... 28 4.3.1 HA Basic Concept ...................................................................................................................................................................... 28 4.3.2 Host Auto-Failover ..................................................................................................................................................................... 30 4.3.3 Backups (Data Backup and Recovery Solution) ................................................................................................................... 32 4.3.3.1 Shared NAS Backups ............................................................................................................................................................. 32 4.3.3.2 3rd Party Backup Soft ware Backups .................................................................................................................................... 33 4.3.4 Disaster Recovery....................................................................................................................................................................... 35 4.4 Networking ..................................................................................................................................................................................... 38 4.4.1 FusionCube E9000 ..................................................................................................................................................................... 38 4.4.1.1 Logical Network Design........................................................................................................................................................ 38 4.4.1.2 Network Interconnection ....................................................................................................................................................... 39 4.4.1.3 Po rt and Bond Solution .......................................................................................................................................................... 40 4.4.1.4 Networking Feature ................................................................................................................................................................ 42 4.4.2 Fusioncube RH8100 v 3 ............................................................................................................................................................. 42 4.4.2.1 Logical Network Design........................................................................................................................................................ 42 4.4.2.2 Network Interconnection ....................................................................................................................................................... 43 4.5 Capacity Expansion and Reduction ............................................................................................................................................ 43 4.5.1 Capacity Expansion.................................................................................................................................................................... 43 4.5.2 Capacity Reduction .................................................................................................................................................................... 44 4.6 Other So lutions............................................................................................................................................................................... 45 4.6.1 Unified Management So lution ................................................................................................................................................. 45 4.6.2 Mult iple SID deploy ment.......................................................................................................................................................... 45 4.6.3 Cluster split and consolidation ................................................................................................................................................. 45
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
iv
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
Contents
4.6.4 Virtualizat ion Solution ............................................................................................................................................................... 46 4.6.5 Co mpatibility............................................................................................................................................................................... 46
5 Specifications of FusionCube for SAP HANA Appliance .................................................. 47 5.1 Delivery Scope of FusionCube for SAP HANA Appliance ................................................................................................... 47 5.2 Specifications of FusionCube for SAP HANA Appliance ...................................................................................................... 48 5.3 Performance of FusionCube for SAP HANA Appliance ........................................................................................................ 52
6 Configurations of FusionCube for SAP HANA Appliance ................................................ 53 6.1 Configurat ion Principles of FusionCube for SAP HANA Appliance ................................................................................... 53 6.2 Configurat ion Calculation of FusionCube for SAP HANA Appliance ................................................................................ 56 6.2.1 Input Informat ion........................................................................................................................................................................ 56 6.2.2 Calculating nu mber of Co mputing Nodes.............................................................................................................................. 56 6.2.3 Calculating NA S Storage space ............................................................................................................................................... 57 6.2.4 Calculating nu mber of Storage Nodes .................................................................................................................................... 59 6.2.5 Calculation of Networking Connections ................................................................................................................................ 63 6.3 Quick Reference of FusionCube Configurations for SAP HANA Appliance ..................................................................... 65 6.3.1 Typical and Reco mmended Configurations ........................................................................................................................... 65 6.3.2 Configurat ion by Memory – NAS For Shared Volu me ........................................................................................................ 69 6.3.2.1 Based on the Ivybridge 1TB nodes: All-SAS d isks. ......................................................................................................... 69 6.3.2.2 Based on the Ivybridge 1TB nodes: composite storage ................................................................................................... 71 6.3.2.3 Based on the Ivybridge 2TB nodes: All-SAS d isks. ......................................................................................................... 73 6.3.3 Configurat ion by Memory – FusionStorage NFS For Shared Volu me .............................................................................. 76 6.3.3.1 Based on the Westmere 512GB nodes: All-SSD ............................................................................................................... 77 6.3.3.2 Based on the Ivybridge 1TB nodes: All-SAS d isks. ......................................................................................................... 77 6.3.3.3 Based on the Ivybridge 1TB nodes: composite storage ................................................................................................... 78 6.3.4 Configurat ion by the Number of Subracks ............................................................................................................................ 79 6.3.4.1 NAS fo r Shared volu me ......................................................................................................................................................... 79 6.3.4.2 Fusionstorage NFS file System for Shared volu me .......................................................................................................... 80
7 Roadmap of FusionCube for SAP HANA Appliance .......................................................... 81 8 Acronyms and Abbreviations ................................................................................................... 82
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
v
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
1
1 Overview of SAP HANA
Overview of SAP HANA The digital technology is sweeping the world. A powerful innovative device is required to process huge data at a fast speed and in real time. SAP HANA appliance is such a device.
1.1 What Is SAP HANA SAP HANA is an in-memory database software platform promoted by the SAP company in June 2011.
SAP HANA is an in-memory database platform provided by SAP.
SAP provides various boundary applications, including data replication and Business Intelligence (BI).
SAP HANA appliance is delivered consisting of hardware, software, and management. SAP HANA appliance is delivered by hardware suppliers who have passed certification of SAP.
SAP HANA has the following characteristics:
Applies to both OLAP and OLTP applications.
Provides the structured query language (SQL), multidimensional expressions (MDX), and business intelligence consumer services (BICS) interfaces for application layers.
Provides multiple data import modes. Is a product with the fastest growth speed in SAP history.
Is the database embracing the fastest data processing speed in the world. SAP HANA is hundreds of and even ten thousands of times faster than other databases according to data from more than 2000 customers who have deployed SAP HANA databases.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
1
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
1 Overview of SAP HANA
Figure 1-1 SAP HANA architecture
1.2 Application Scenario of SAP HANA Figure 1-2 Application scenario
SAP HANA can function as the following:
Accelerator for other databases
Data source for report analysis OLAP application database
OLTP application database
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
2
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
1 Overview of SAP HANA
The unified database platform of all SAP applications
1.3 Systems Interworking of SAP HANA SAP HANA appliance stores, queries, and calculates data of the in-memory database. The following figure shows system interworking with SAP HANA appliance. Figure 1-3 Systems interworking with SAP HANA
Data Sources and Data Services: Provide analysis data for SAP HANA appliance. Some data (from SAP ERP system) can be exported into SAP HANA through the replication agent. Some data (from SAP BW system and third-party system) must be converted using the Data Services software and then is exported into SAP HANA appliance.
Admin Workstations: The client of SAP HANA system administrator. It can manage and use SAP HANA appliance through some software, such as SAP HANA Studio software. End Users Workstations: The user client of the SAP HANA system. Terminal users can use the SAP HANA system through excel and web BI.
SAP HANA Database client: Provides interfaces, such as JDBC, ODBC, ODBO, and SQL DBC so that each type of software can access SAP HANA appliance using the standard SQL mode.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
3
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
1 Overview of SAP HANA
1.4 Logical Function Architecture of SAP HANA Figure 1-4 Software architecture of SAP HANA
SAP HANA Database is core of the entire SAP system. Modules highlighted in a red box are main modules by function.
Session Management: Setup the connects to various clients and manages the sessions between HANA and clients. Replication Server and Load Controller: Replicates data of such as ERP/SAP NetWeaver/third-party database into the SAP HANA database. Transaction Manager: Manages transactions of the database. After one transaction has been finished, the data persistence is required. Authorization Manager: Manages authorizations for clients connected to the SAP HANA database.
Metadata Manager: Cooperates with Modeling and SAP Data Service (an ETL service software of the SAP) in SAP HANA Studio to manage metadata established in or exported into the SAP HANA database.
SQL Parser: Analyzes and parses SQL requests from service clients. It also divides one SQL task into multiple subtasks and distributes them to other nodes for processing.
MDX: The multidimensional expressions. It analyzes and parses complicated computing requests from applications, such as excel.
SQL Script: Provides the complicated programming capability similar to the storage procedure.
Calc Engine: Performs the SQL calculation.
ROW Store: Stores the in-memory database.
Column Store: Stores the in-memory database.
Page Management and Logger: The data persistence processing module. It writes data in the in-memory database into the SSD or HDD in synchronous or asynchronous mode.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
4
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
1 Overview of SAP HANA
1.5 Product Portfolio of SAP HANA From the view of user buying, SAP HANA product can be divided into two types:
Appliance: It means that HANA database software, certified HANA hardware and management software are packaged into a unified product to be sold to customers.
TDI: Tailored Data Center Integration. Customers can select to buy HANA servers or storages freely.Certified HANA support engineers will be responsible for installing the whole system.
From the view of node scaling mode, SAP HANA can also be divided into two types:
Scale Up System: It is also called single node system. There is only an active node in this system. Such a system only has vertical scaling ability. When database data volume becomes more and more, we enlarge the system by adding more CPU/Memory/Disk.
Scale Out System: It is also called multi-node system or cluster system. There are many nodes in such a HANA system. The system scaling mode is mainly horizontal scaling(It means that more nodes can be added into the system).
Combining the previous two types, we can get more SAP HANA product portfolio: Single node appliance system: A single node system which is bought and delivered with a means of appliance. Multi node appliance system: A multi-node system which is bought and delivered with a means of appliance. Multi node TDI system: A multi-node/cluster system which is bought and delivered with a means of TDI.
The following chapters of this document describe SAP HANA appliance cluster system.
1.6 Overview of SAP HANA Appliance Cluster System 1.6.1 Main Concepts of SAP HANA Appliance Cluster System
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
5
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
1 Overview of SAP HANA
Figure 1-5 Main Concepts of SAP HANA Cluster System
As shown in the above figure, A SAP HANA cluster system is made of many nodes.
In this document we have the following concepts about node name: SAP HANA Node: It is a logical unit which composes a SAP HANA cluster system. o
One SAP HANA node includes CPU, memory, Log Volume and Data Volume with specified capacity.
o
A SAP HANA cluster system will be charged with size and amount of HANA node when it is sold.
o
The physical form of a HANA node can be various. It depends on the detail implement of hardware partners.
Computing Node: It is responsible for in-memory database computing. Storage Node: It is responsible for in-memory database persistent storage. A SAP HANA node must include a computing node, but it may not include an independent storage node physically. So in some extent a SAP HANA node equates a computing node.
A SAP HANA node can be different role according to its function in the cluster system. It can be a master node, slave node, or standby node, which cannot be co-located in physical deployment.
Master node: The configuration role (as shown in the following ,Name Server Config Role) of three nodes can be the master node, but only one master node is activated(At that time, it means that actual role of this node is master). Which one is the activated
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
6
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
1 Overview of SAP HANA
node is determined by the management mechanism of the SAP HANA software. The activated master node coordinates global transactions and stores global metadata of the whole cluster system.
Slave node: Multiple slave nodes are configured in a SAP HANA cluster system. The slave node in cluster system buffers required metadata and performs database operations assigned by the master node. One cluster contains multiple slave nodes.
Standby node: Takes over services from the faulty node in the event of failover. No or multiple standby nodes can be configured in one cluster. In common situations, software of the standby node is running but no data is stored in its in-memory database, and the standby node cannot process any services. The failover occurs automatically.
Figure 1-6 Roles of the SAP HANA node
1.6.2 Software Deployment of SAP HANA Appliance Cluster System Figure 1-7 Deployment of SAP HANA main software processes
The following software processes are installed on the master node:
SAP HANA Studio: As the management client of the SAP HANA database, it implements modeling establishment, configuration, monitoring, and alarm generation functions for the SAP HANA database.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
7
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
1 Overview of SAP HANA
SAP HANA Client: As the database client software, it provides interfaces, such as ODBC for users to access database data.
SAP HANA Database: The software implements in-memory database functions. It is the core software package in a SAP HANA system. SAP SMD Agent: As an agent it will integrate SAP HANA into a whole SAP solution.
SUM for SAP HANA: It is the software upgrade manager of SAP HANA. The master node automatically connects to the software update platform through this software. This process upgrades SAP HANA software on all master, slave, and standby nodes. The OS and drives are installed and upgraded by hardware suppliers.
SAP Host Agent: Restarts or stops nodes as instructed by remote commands. These operations are performed on SAP HANA software.
SAPCAR: Installation scripts which take effect only for SAP HANA software.
The following software processes are installed on the slave or standby node:
SAP Host Agent
SAP HANA Client SAP HANA Database
SAP SMD Agent
As the core software package of a SAP HANA cluster system, SAP HANA database software consists of the following some processes as shown in Figure 1-8 Figure 1-8 Main processes of the SAP HANA database software
Daemon: Monitoring process. Monitoring the health state of other SAP HANA database processes.
NameServer: The cluster management process. It is responsible for collecting the information of cluster system and controlling the takeover of cluster system.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
8
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
1 Overview of SAP HANA
Preprocessor: The process of
IndexServer: The core process that performs database operations. It is responsible for storing database data and executing data computing.
StatisticsServer: statistics process. It records running information of the database. This process is running only on the master node.
Xsengine: HTTP service process. Clients can connect to SAP HANA database directly in a interface of HTTP/HTTPs.
SQL statement preprocess.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
9
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
2
2 Overview of Huawei SAP HANA Product
Overview of Huawei SAP HANA Product
2.1 Product Portfolio As the global strategic partner of SAP company, Huawei can provide all kinds of SAP HANA products and solutions, including:
Single node SAP HANA solution: It is implemented based on the RH5885/RH8100 rack server. Huawei has passed certification for all-series (including XS/S/M/L/L+ series) SAP HANA single mode edition.
Scale out SAP HANA solution: It can be implemented based on Huawei rack servers (RH5885/RH8100), Huawei blade servers ( E9000), Huawei distributed storage (FusionStorage) or Huawei enterprise-class storage (OceaStor) . The scale out edition has two deployment modes:
Appliance Solution: Customers buy SAP HANA integrated device including all of hardware, software and management.
Converged infrastructure solution:
E9000+FusionStorage cluster scenario(support 1-16 nodes);
RH8100v3+FusionStorage cluster scenario(support 1-16 nodes);
SAN storage solution: RH5885+ S5500T IP SAN device.
This solution is mainly used for the small cluster scenario(1-4 nodes).
TDI Solution: Customers buy either servers of SAP HANA or storage independently and setup SAP HANA system by professional SAP HANA engineers. The system consists of two parts.
HANA Server: Customer can buy HANA servers from Huawei or other suppliers. If from Huawei they are Huawei FusionServer servers: RH5885H V3,RH8100 V3,E9000 V3.
HANA persistent storage: Customer can buy HANA persistent storage from Huawei or other suppliers. If from Huawei they are Huawei enterprise storage: S5500T,S5600T,S5800T,S6800T,18500,18800 and 18800F.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
1
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
2 Overview of Huawei SAP HANA Product
2.2 Product License Huawei solution for SAP HANA appliance is sold mainly by ISV&SI, It means, the channel distributor or retailer signs contracts with customers about Huawei solution for SAP HANA appliance in most scenarios. In some important scenarios, Huawei can also sign contracts with customers directly.
Database license: The SAP HANA database software license is sold with 64 GB memory as the unit. Customers need buy license of SAP HANA database from SAP.
Infrastructure license: The hardware infrastructure license is sold by hardware suppliers. According to the trade-off agreed between Huawei and SAP, What Huawei provides includes: Hardware devices (computing, storage, and network), hardware device services and OSs. In converged infra scale out edition, FusionStorage software license will be charged based on disk storage capacity in unit of TB.
2.3 Scope of this document For details of the single node edition, see Huawei SAP HANA Appliance Technical White Paper -Single Node Edition . For details of SAN storage-based scale out edition: see Huawei SAP HANA Appliance Technical White Paper-Scale Out Edition (SAN Storage). The following chapters of this document describe FusionCube for SAP HANA appliance based on converged infrastructure. It describes architecture, hardware,software, configuration and roadmap about Huawei SAP HANA appliance.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
2
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
3 System Composition of Huawei FusionCube for SAP HANA Appliance
3
System Composition of Huawei FusionCube for SAP HANA Appliance 3.1 Overall Architecture Figure 3-1 shows the architecture of Huawei FusionCube solution for SAP HANA appliance. Figure 3-1 Architecture of Huawei FusionCube for SAP HANA
Huawei FusionCube solution for SAP HANA appliance is based on FusionCube Converged Infrastructure, which is a flagship product of Huawei IT product line.
Hardware device: The E9000 is used to support modular design for components, such as computing, storage, switching, power supply, and fans. Computing, storage, and network are deployed in the E9000 subrack. Computing and storage nodes can be flexibly configured
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
3
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
3 System Composition of Huawei FusionCube for SAP HANA Appliance
as required. Various switch modules are supported, such as GE, 10GE, FC, FCoE, and IB modules. It’s also support large memory capacity server RH8100v3.
FusionStorage: uses distributed cloud storage technologies to provide high-performance high-reliability block storage services by scheduling local hard disks on servers in an optimized manner.
FusionCube Center: It provides a unified web interface for users to perform service operations, including user management, cluster management, VM management, unified hardware management, alarm reporting, and system monitoring.
File system: Log/Data Volume: Huawei uses XFS local file system and FusionStorage distributed block storage to provide low latency, high bandwidth persistent storage.
Shared Volume: Huawei can support fusionstorage nfs file system.
3.2 Converged Devices for Computing, Storage and Network 3.2.1 E9000 3.2.1.1 E9000 High-Density Subrack Huawei FusionCube for SAP HANA uses Huawei Tecal E9000 (short for E9000) as the hardware. The E9000 with 12-U height is an integrated hardware device for computing, storage, and network. Huawei E9000-series computing nodes, storage nodes, service processing nodes, and corresponding resource expansion nodes can be deployed in the E9000.
Houses eight horizontal full-width computing nodes or 16 half-width computing nodes and supports slot division. Full-width blades are used in all nodes of Huawei FusionCube for SAP HANA.
Supports 850 W heat dissipation for half-width slots. Supports 1700 W heat dissipation for full-width slots. Supports a maximum of 2 CPUs and 24 DIMMs in half-width slots.
Supports a maximum of 4 CPUs+48 DIMMs in full-width slots.
Supports 32 CPUs/480 cores on a single chassis, with a maximum of 12 TB memory.
Supports a maximum of 15.6 Tbit/s switching capacity on the backplane.
Provides four (two pairs) switching slots, supports multiple types of switching protocols, such as Ethernet, Fibre Channel (FC), Fibre channel over Ethernet (FCoE), and Inifiband(IB), and supports straight-out I/O interfaces.
A FusionCube cabinet can house three E9000 subracks.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
4
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
3 System Composition of Huawei FusionCube for SAP HANA Appliance
Figure 3-2 Appearance of E9000
3.2.1.2 High-Density Blade of E9000 The FusionCube supports multiple types of blades.
3.2.1.2.1 Ivybridge EX-based CH242 V3 Computing Blade This computing blade has the same structure as the CH242 computing blade. The differences lie in the following:
The CH242 computing blade supports the Westmere platform-based CPU. While the CH242 V3 computing blade supports the CPU based on the Ivybridge platform and each CPU can have up to 15 cores.
The CH242 computing blade supports only one MZ512 10GE NIC. While the CH242 V3 computing blade supports two MZ512 10GE NICs.
The motherboard of the CH242 computing blade carries an RAID card. While the motherboard of the CH242 V3 computing blade does not carry the RAID card and must be configured as required.
3.2.1.2.2 CH222 Management Blade Figure 3-3 CH222 management blade
The CH222 management blade has the following features:
Uses 2-CPU Intel® Xeon® E5-2600 V1/V2 processors with high performance.
Provides 24 DIMM slots and up to 768 GB DDR3 memory.
Supports a maximum of fifteen 2.5-inch SSD, SAS, or SATA disks.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
5
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
3 System Composition of Huawei FusionCube for SAP HANA Appliance
Uses hot swappable architecture, supports in-service replacement of a single hard disk, and provides one full-height 3/4-length PCIe x16 slots.
When CH222 functions as the management blade, only two hard disks are installed as the system disks.
3.2.1.2.3 CH222 Storage Blade The CH222 storage blade has the same parameters as the CH222 management blade. While hardware disks are different during the actual usage. In addition to two system disks, another 12 or 13 hard disks are installed on the CH222 storage blade.
3.2.1.2.4 CH223 Storage Blade Figure 3-4 CH223 storage blade
Main parameters are as follows: Form
A full-width 2-socket server blade with one slot
Number of processors
1/2
Supported processor types
Intel Xeon E5-2600 Intel Xeon E5-2600 V2
Number of memory slots
24 DDR3 DIMM slots, providing a maximum memory of 768 GB
Number of hard disks
Two 2.5-inch SSD, SAS or SATA hard disks
Supported RAID
RAID 0 and RAID 1
PCIe expansion
Two PCIe x16MEZZ daughter cards (CH220/CH221) can be added (MEZZ2 is occupied). Four PCIe x8FHHL standard cards (CH220) or two PCIe x16FHHL standard card (CH221) can be added.
Supported OS types
Microsoft Windows Server 2008/2012 Red Hat Enterprise Linux SUSE Linux Enterprise Server Citrix XenServer VMware ESX
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
6
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
3 System Composition of Huawei FusionCube for SAP HANA Appliance
Operating temperature
5C40C
Dimensions (H x W x D)
60.46 mm x 525 mm x 423 mm
Key characteristics:
Powerful expansion: supports standard PCIe expansion slots. Up to four PCIe cards can be configured for a full-width slot and each PCIe card can be inserted with a 2.4 TB SSD.
1.5 times height of DIMMs: The cost-effectiveness leads the industries.
3.2.1.3 High-Performance Switching of E9000 3.2.1.3.1 Ethernet Switch Huawei FusionCube for SAN HANA uses the CX310 switch module as the Ethernet switch. CX310, the switching and control unit of the E9000 blade server, provides the data switching function for each service slot of the entire system and provides external service interfaces and management interfaces. Appearance of the CX310 is as follows: Figure 3-5 Appearance of the CX310
Performance indicators of the CX310 are as follows: Table 3-1 CX310 performance indicators Item
Description
Number of ports on the panel
One BMC serial port and one SYS serial port
Sixteen 10GE uplink Ethernet optical ports
Switching capacity
1.28 Tbit/s
Packet forwarding rate
960 Mpps
Ethernet
Supports full-duplex and autonegotiation working modes.
Supports GE and 10GE rates.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
7
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
3 System Composition of Huawei FusionCube for SAP HANA Appliance
Item
Description
VLAN
Supports access, trunk, and hybrid modes.
Supports default VLANs.
Supports QinQ and enhanced selective QinQ.
Supports MUX VLANs.
Supports flow classification based on the combination of layer-2, layer-3, layer-4, and 802.1p protocols.
Supports ACLs, committed access rate (CAR), remark, and traffic scheduling.
Supports various queuing modes, such as priority queue (PQ), weighted round robin (WRR), deficit round robin (DRR), PQ+WRR, and PQ+DRR. Supports anti-congestion mechanisms, such as weighted random early detection (WRED) and tail drop.
QoS
Security and management
Supports traffic shaping (TS).
Provides hierarchical protection using commands, preventing unauthorized access.
Defends against Denial of Service (DoS), Address Resolution Protocol (ARP), and Internet Control Message Protocol (ICMP) attacks.
Supports port isolation, port security, and sticky MAC.
Supports binding between IP addresses, MAC addresses, ports, and VLANs. Provides various authentication modes, such as Authentication, Authorization and Accounting (AAA), Remote Authentication Dial In User Service (RADIUS), and Huawei Terminal Access Controller Access Control System (HWTACACS). Supports remote network monitoring (RMON).
3.2.1.3.2 FC Switching Huawei SAP HANA appliance can provide FC interface for HANA FC-interface based backup/recovery. In the E9000, It is CX311 switch module that provides FC capability.CX311 is similar to CX310, but it has an extra FC sub board. CX311 can provide up to 8 FC interfaces.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
8
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
3 System Composition of Huawei FusionCube for SAP HANA Appliance
Figure 3-6 Appearance of the CX311
3.2.1.3.3 IB Switching The HANA appliance uses the CX610 switch module as the storage network to implement the InfiniBand (IB) switching. Two CX610 switch modules are configured for each subrack. The following figure shows the physical appearance of the CX610. Figure 3-7 Appearance of the CX610
Provides one BMC serial port and one SYS serial port. Provides 18 QSFP+ optical ports for connecting external devices.
3.2.2 RH8100 V3 HANA hardware using the next-generation product Huawei RH8100 V3. RH8100 V3 servers using Intel E7-8800 V3 series CPU, it support large memory capacity Key services are critical to enterprises and require highly reliable servers. To meet performance requirements of real-time data analysis in the Big Data era, in-memory databases need to provide large memory capacity, high computing capability, and good scalability. Some other applications, such as heavy-load virtualization applications and high-performance computing (HPC), also require high-performance, high-reliability servers. The RH8100 V3 has high reliability, outstanding performance, a leading architecture, and an open and easy-to-use management platform, suited for mission-critical services, in-memory databases, virtualization, and HPC.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
9
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
3 System Composition of Huawei FusionCube for SAP HANA Appliance
Figure 3-8 HANA 服务器 RH8100 V3
3.2.2.2 Product Features Performance and Scalability The RH8100 V3 supports the following features to improve performance and scalability while reducing the total cost of ownership (TCO):
Supports a maximum of eight processors (four processors can also be configured.) − Each Intel® Xeon® E7-8800 v2 processor ensures high processing performance by providing up to 15 cores, one L3 cache of 37.5 MB, and three 8.0 GT/s QPI links between processors. −
Each Intel® Xeon® E7-8800 v3 processor ensures high processing performance by providing up to 18 cores, one L3 cache of 45 MB, and three 9.6 GT/s QPI links between processors.
Uses Intel Turbo Boost Technology, which allows processor cores to run faster than the frequency specified in the TDP configuration if they are operating below power, current, and temperature specification limits.
Uses Intel Hyper-Threading Technology, which enables each processor core to run up to two threads, improving parallel computation performance.
Uses the hardware-assisted Intel® Virtualization Technology (Intel® VT), which allows OS vendors to better use hardware to address virtualization workloads.
Large memory capacity and high memory bandwidth − −
Provides 18 PCIe slots. −
Ten PCIe slots are on the back I/O module (BM), used for installing standard cards that need to connect to external cables, for example, NICs, HBAs, and IB cards.
−
Six PCIe slots are on FM-A, used for installing standard PCIe SSD cards (Huawei ES3000) or GPUs that do not need to connect to external cables. Up to two compatible GPUs or six standard PCIe SSD cards are supported. Two PCIe slots are dedicated for RAID controller cards.
−
Supports a maximum of 192 DDR3 DIMMs with up to 1600 MT/s bandwidth when the server is configured with E7 v2 processors. Supports a maximum of 192 DDR4 DIMMs with up to 1866 MT/s bandwidth when the server is configured with E7 v3 processors.
Provides two LOMs with GE or 10GE ports. The two LOMs do not occupy standard PCIe slots.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
10
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
3 System Composition of Huawei FusionCube for SAP HANA Appliance
Supports PCIe 3.0, which increases the maximum theoretical I/O bandwidth by 60% compared with PCIe 2.0.
Availability and Serviceability The RH8100 V3 provides 60 RAS features to improve availability and serviceability:
Supports memory migration, combined memory board hot swap, to implement DIMM replacement without shutdown, which is leading in the industry. Complies with component selection criteria for carrier-grade component reliability and derating, and meets carrier-grade reliability test requirements. Implements redundancy for key components, such as the BIOS flash memory, iBMC image, PSUs, and fan modules. Supports faulty DIMM indication on an offline memory board, which allows faulty DIMMs to be indicated on a removed memory board. Supports redundancy and hot swap without opening the chassis cover for key components such as PSUs, fan modules, and hard disks. These features enable quick replacement of faulty components without interrupting system operation.
Supports hot swap of PCIe cards in some PCIe slots without opening the chassis cover, which implements PCIe upgrades and replacement without interrupting system operation.
Supports hot swap of hard disks and uses RAID to protect data on hard disks and maximize system uptime.
Uses the Huawei iBMC to monitor system parameters in real time, trigger alarms, and perform recovery actions in case of failures, minimizing system downtime. Integrates fault management and predictive failure analysis (PFA) to greatly improve fault locating efficiency and reduce unexpected downtime.
Provides a 3.5-inch touch-screen LCD diagnosis panel to facilitate fault locating and shorten the fault recovery time. Provides a three-year warranty for parts replacement and onsite limited repair for the server, and provides 9 x 5 (a 9-hour-a-day, 5-day-a-week) next business day (NBD) support. Optional service upgrades are available.
Manageability and Security The RH8100 V3 provides the following features to simplify local and remote server management:
Provides iBMC, which monitors server operating status and implements remote management in compliance with Intelligent Platform Management Interface (IPMI) 2.0.
Integrates the Unified Extensible Firmware Interface (UEFI)to improve setup, configuration, and update efficiency and simplify fault handling.
Supports the Intel® Advanced Encryption Standard New Instructions (AES NI), which allows faster and stronger encryption.
Provides the Intel Execute Disable Bit (EDB) function, which works with the supported OS to prevent certain types of malicious buffer overflow attacks. Adopts strict security test criteria and the chip-level Huawei iBMC, ensuring secure system operating.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
11
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
3 System Composition of Huawei FusionCube for SAP HANA Appliance
Energy Efficiency The RH8100 V3 provides the following features to reduce energy consumption and operating expense (OPEX) and increase energy efficiency:
Uses E7-8800 v2 series processors to deliver nearly double performance with only 20% higher TDP, compared with the previous-generation product. Uses Huawei Platinum PSUs with a maximum power conversion efficiency of 94%.
Supports power capping when it operates in single-system mode.
Provides Intel® Intelligent Power Capability, which allows a processor to be powered on or off based on site requirements.
Uses low-voltage, low-power E7-8800 v2 processors, suited for data centers and telecommunication environments that have power and thermal limitations.
3.2.2.3 Product Specifications Table 3-2 Product specifications Item
Specifications
Form factor
8U rack server without holding rails
Processor
A maximum of eight Intel® Xeon® E7-8800 v2 (IvyBridge-EX) processors. Each processor supports up to 37.5 MB L3 cache capacity and 15 cores. (Four processors can also be configured.)
A maximum of eight Intel® Xeon® E7-8800 v3 (Haswell-EX) processors. Each processor supports up to 45 MB L3 cache capacity and 18 cores. (Four processors can also be configured.)
Chipset
Intel Patsburg PCH-J (C602-J)
Number of DIMM slots
Up to 192 DIMM slots in 16 memory boards(Each processor supports two memory boards.)
E7 v2: supports only DDR3 DIMMs and memory boards.
E7 v3: supports only DDR4 DIMMs and memory boards.
NOTE 192 DIM M slots are supported only when you use the 12-DIM M memory boards.
Maximum memory capacity
12 TB (using 64 GB DIMMs)
Number of hard disks
Twelve (FM-A), twenty-four (FM-B), or eight (FM-C) 2.5-inch hot-swappable SAS/SATA HDDs or SSDs
RAID support
A RAID controller card supports either of the following:
RAID 0, 1, 10, and 1E
RAID 0, 1, 10, 5, 50, 6, and 60, a maximum cache capacity of 2 GB, and a supercapacitor for power-off protection One or two RAID controller cards of the same model
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
12
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
3 System Composition of Huawei FusionCube for SAP HANA Appliance
Item
Specifications
Network port
The LOM can be flexibly configured to provide either of the following ports:
PCIe slot
Up to two LOMs
2-port GE-RJ45/4-port GE-RJ45/2-port 10GE SFP+/2-port 10GE RJ45
Supports a maximum of 18 PCIe slots if the server is configured with FM-A.
Two PCIe 3.0 x16 and eight PCIe 3.0 x8 slots are on the BM, used for installing standard cards that need to connect to external cables, for example, NICs, HBAs, and IB cards.
Two PCIe 3.0 x16 and four PCIe 3.0 x8 slots are on FM-A by using two PRMs, used for installing standard cards that do not need to connect to external cables, for example: Six Huawei ES3000 PCIe SSD cards Two Huawei-certified GPUs Three Huawei ES3000s and one GPU
External ports and buttons
Two PCIe slots are dedicated for RAID controller cards.
Front panel: two USB 2.0 ports, one Power button, one UID button, one VGA port, and one touchable LCD diagnosis panel
Rear panel: Two consoles. Each provides two USB 2.0 ports, one VGA port, one serial port, one GE management network port, one UID indicator, and one LOM slot.
NOTE Some ports on the two management modules on the rear are unavailable. For details, see 错误!未找到引用源。错误!未找到引用 源。.
DVD-ROM drive
DVD-RW drive with one SATA port
Graphics card
The graphics card is integrated into an HFC, and supports 32 MB display memory and a maximum resolution of 1280 x 1024.
PSU
You can choose 3000 W AC, 2000 W AC, or 2500 W DC PSUs. The PSUs can be configured in 2+2 or 1+1 redundancy mode.
System management
Supports remote management, WebUIs, virtual KVM, and standard protocols such as IPMI 2.0 and SNMP.
Security
Power-on password and administrator password
Supported OSs
Windows Server 2012 R2, RHEL 6.5, SLES 11.3, and so on
Warranty
Huawei provides a three-year warranty for parts replacement and onsite limited repair for the server. Huawei provides 9 x 5 (a 9-hour-a-day, 5-day-a-week) next business day (NBD) support. Optional service upgrades are available.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
13
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
3 System Composition of Huawei FusionCube for SAP HANA Appliance
Item
Specifications
Dimensions (H x W x D)
352 mm x 447 mm x 855 mm (13.86 in. x 17.60 in. x 33.66 in.)
Weight
Maximum weight: 160 kg (352.80 lb)
Physical environment
Operating environment
Ambient temperature: 5°C to 40°C (41°F to 104°F) The operating temperature decreases by 1°C (1.8°F) as the altitude increases by 300 m (984.25 ft). Maximum altitude: 3000 m (9842.52 ft)
Ambient humidity: 10% RH to 90% RH (twmax = 29°C)
Storage environment
Acoustic noise
Ambient temperature: –40°C to +65°C (–40°F to +149°F)
Ambient humidity: 5% RH to 95% RH (twmax = 38°C)
The data listed in the following is the declared A-weighted sound power levels (LWAd) and declared average bystander position A-weighted sound pressure levels (LpAm) when the server is operating in a 23°C (73.4°F) ambient environment. Noise emissions are measured in accordance with ISO 7999 (ECMA 74) and declared in accordance with ISO 9296 (ECMA 109).
Idle: LWAd: 6.6 Bels LpAm: 50.3 dBA
Operating: LWAd: 6.9 Bels LpAm: 52.7 dBA
NOTE The actual sound levels generated during server operating vary depending on the server configuration, loads, and ambient temperature.
3.3 Shared Storage Device In Huawei FusionCube solution for SAP HANA appliance NFS is used for Shared volume of SAP HANA. Huawei has its own NAS device to finish NFS function, it is N8500.
3.3.1 Overview of NAS from Huawei 3.3.1.1 Overview of product Oceanstor N8000 cluster NAS storage system from Huawei is a leading unified storage platform for middle-end and high-end application. It can meet the business requirement of the data management with high performance, high scaling and high-efficiency and unified storage
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
14
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
3 System Composition of Huawei FusionCube for SAP HANA Appliance
from all kinds of industries: e.g. carrier, digital media, high-performance computing, government and R&D Figure 3-9 Appearance of the N8500
The upper figure is the rack deployment of N8500. In the minimum configuration of N8500 there is a file engine chassis, a disk controller chassis and a disk chassis.
3.3.1.2 Characteristics of product High Performance
Leading performance：In SPEC sfs2008 benchmark test there is a leading performance. In scenario of the NFS protocol there is output of 3,064,602 OPS。
Cluster Architecture: there can be many file engines in a NAS system and all file engines can work actively and concurrently.
Global shared storage space: all computing node can access the same storage space simultaneously; File engine has a capability of load balance.
High Availability
No single of failure: all components are redundant, there is no single of failure in N8000 system.
High Scalability
Engine Node scaling: The Engine nodes can scale out into 24 nodes with smooth, the system performance capability can enlarge with adding of engine number.
Storage space extending: Storage space can be extended into up to 15PB
File system ability: There can be up to 256TB space and up to 1 billion files in a file system
High-Efficiency Data Management
Dynamic storage tiering: Based on file access frequency and time policy storage can be moved among different storage media. This movement is no effect to applications, so it will reduce the customer CAPEX.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
15
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
3 System Composition of Huawei FusionCube for SAP HANA Appliance
File system mapping: Several replicas can be stored in the same storage system by setting file system mapping, which will improve the reliability of online data.
Agile Space Management: User storage space can be allocated and used with great agility.
Domain Management: Support AD/NIS/LDAP
Backup/Recovery: Symantec NetBackup Client can be installed in the N8500 file engine node and this will reduce the time of backup and get good backup performance.
Unified Storage
Unified Protocols：All kinds of storage network protocol like iSCSI、FC、FCoE、NFS、 CIFS、HTTP、FTP etc can be supported. The protocols from SAN and NAS can be supported at the same time. Structure data and unstructured data can be supported at the same time.
Unified Management: File storage and block storage can be managed in a unified, simple GUI.
Device Monitoring:
In GUI state and statistics information for real-time /history data
can be displayed and monitored. Alarms can be shown in many modes like voice, light and e-mail to master the system health state at any time.
File-grade data protect technology: multi-layer data protection can be adopted in a file system grade, e.g. file system snapshot, snapshot recovery and remote replication.
Block-grade data protect technology: Smart Cache dynamic data caching technology and LUN based snapshot, copying and synchronous/asynchronous remote replication add-on software features can be used during block-grade data protection.
3.3.1.3 Specification of product
Specification indicators of the N8500 are as follows: Table 3-3 N8500 main specification indicators Sort System
Sub Sort
Item
Description
System architecture Number of engines
Multi-node active-active cluster architecture 2 to 24
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
16
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
Sort
Sub Sort
Item
Cache per node
GE per node 10GE per node Disk type storage capacity Maximum capacity of a file system
performance
Protocol
Software feature
Peak consumption
Dimensions
Basic: 16 GB; Standard: 24 GB Enterprise: 48 GB; Enhanced: 96 GB (standard configuration) and scalable to 192 GB 4*GE,up to 16*GE Up to 6*10GE SSD, SAS, NL-SAS, SATA, and intermixing of different disks Scalable, up to 15 PB Scalable, up to 256 TB 256
Maximum number of files
200 million files per file system
A folder supports a maximum of 32,000 sub-folders
Maximum number of file system snapshots
Not more than 128 periodic snapshots per file system
Not more than 512 snapshots per file system
Maximum number of remote replication
32
Maximum number of file system mirrors
3
Bandwidth of sequence read/write (NFS/CIFS/FTP/HTTP)
≥ 300 MB/s (single node)
≥ 500 MB/s (single node with the 10 Gbit/s network card)
OPS (NFS/CIFS)
≥ 50000 (for one node), [OPS: Operations per second]
Storage Protocol Access Control
NFS, CIFS, FTP, HTTP, FCP, FCoE, iSCSI NIS，Microsoft Active Directory，LDAP
Other protocol
NDMP，SNMP，NTP
RAID level
RAID 0，1，3，5，6，10，50
OS
DST, LoadBalance, Snapshot, Mirror, Quota, Replication, WORM, SmartCache, HyperThin, HyperImage, HyperCopy, HyperClone, and HyperMirror GUI, CLI Windows，Linux，Mac OS，Solaris，AIX，HP-UX
Alarm mode
E-mail，SNMP，Syslog
Range of the rated power input voltage
AC：100V～127V/200V～240V
File Engine Enclosure
<=830W
Disk Controller Enclosure Disk Enclosure
<=539W (S5500T) <=307W (2U SAS ) 4U(2 nodes)，175mm(H) * 446mm(W) * 502mm(D)
Management mode
Power input
Description
Maximum number of file systems
Value-added software
Physical Feature
3 System Composition of Huawei FusionCube for SAP HANA Appliance
File Engine Enclosure
DC：–48V～–60V
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
17
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
Sort
Sub Sort
Item Disk Controller Enclosure Disk Enclosure Rack
Weight
File Engine Enclosure Disk Controller Enclosure Disk Enclosure
3 System Composition of Huawei FusionCube for SAP HANA Appliance
Description 2U: 86.1mm(H) % 446mm(W) % 582mm(D) 4U: 175mm(H) % 446mm(W) % 502mm(D) 2U,86.1mm(H) % 446mm(W) % 412mm(D) 4U,175mm(H) % 446mm(W) % 412mm(D) 42U， 205cm(H) % 57.5cm(W) % 118cm(D) 43.6 kg in 2 nodes 2U, ≤29.9Kg/per enclosure in full configuration 4U, ≤43.6Kg/per enclosure in full configuration 2U,≤26.5Kg/per enclosure in full configuration 4U,≤42.64Kg/per enclosure in full configuration
3.3.2 NAS in FusionCube SAP HANA appliance 3.3.2.1 Main Compenonts In FusionCube solution for SAP HANA appliance the used NAS components are as follows: NAS File engine Disk controller enclosure Disk enclosure
Figure 3-10 N8500 in FusionCube SAP HANA cluster appliance
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
18
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
3 System Composition of Huawei FusionCube for SAP HANA Appliance
The NAS system provides the following outer interface: Business Line: Two units of the NAS file engine provide 4*10GE for connecting to SAP HANA nodes. This will assure the network bandwidth and no single point failure of business network. Management Line: Two units of the NAS file engine and two units of disk controller enclosure provide 4*GE for system management. This will also assure no single point failure of management network.
3.3.2.2 Main Configurations Main configurations of the N8500 in FusionCube SAP HANA cluster appliance are as follows: Table 3-4 N8500 main configurations in FusionCube SAP HANA cluster appliance File Engine Enclosure CPU Memory Data Disk Business Network Management Network Storage Network OS Software Height
Disk Enclosure
2CPU 48G / 4*10GE 2*GE 8*8GB FC SLES11.1 NFS Server 4U NFS, CIFS, FTP, HTTP, FCP, FCoE, iSCSI, NDMP LAN-Free
2CPU 48G / / 2*GE 8*8GB FC / / 4U
/ / 24*900G 10K SAS / / / / / 2U
/
/
Disk Type
/
/
RAID Level
RIAD5,RAID6,RAID10
/
Protocol
3.4
Disk Controller Enclosure
SSD, SAS, NL-SAS, SATA /
High-Performance Distributed Storage Sub-System Huawei FusionCube solution for SAP HANA appliance uses FusionStorage distributed block storage subsystem developed by Huawei as the storage software. High-performance becomes a key requirement on enterprise storage systems. However, traditional storage systems cannot meet the demanding storage requirements of increasing data by capacity expansion. Moreover, high-end storage devices that serve as independent NEs are not only expensive but also need dedicate management and maintenance.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
19
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
3 System Composition of Huawei FusionCube for SAP HANA Appliance
With its high performance, Huawei-developed distributed storage software, FusionStorage, integrates computing and storage and combines local disks on all servers into a virtual storage resource pool. Therefore, FusionStorage can completely take the place of external SAN device in some scenarios. The integration of computing and storage in FusionStorage achieves high performance, high reliability, and high cost-effectiveness.
Innovative architecture and design: The innovative architecture empowers the FusionStorage to achieve optimal performance. This performance optimization allows you to fully use system resources and enables the automatic workload allocation among all hard disk drives. In addition, FusionStorage supports a series of advanced functions, such as thin provisioning and snapshot, without compromising deteriorating system performance.
Consistent and predicable performance expansion: The load balancing among all disks and distributed cache structure of the FusionStorage allow you to expand system performance linearly and smoothly by adding server nodes. FusionStorage ensures consistent high performance at peak and off-peak hours regardless of the changes of the service amount and snapshot mode or even though the component becomes faulty.
Flexibility and self-recovery: The FusionStorage retains high flexibility and runs properly even if a hardware device becomes faulty. In addition, the self-recovery capability empowers FusionStorage to defend against more hardware faults that may occur after its initial recovery from failures.
Deep integration of computing and storage: The FusionStorage integrates the local disks on all servers as a virtual storage resource pool and delivers automatic management, high running performance, and out-of-the-box packaging for quick deployment. The FusionStorage does not serve as an independent NE for dedicate configuration or management. In this way, resources on the computing nodes can be fully used, which not only provides users with high storage performance, but also reduces costs.
3.4.1 High-Performance Snapshots FusionStorage provides a snapshot mechanism, which allows the system to capture the status of the data written into a logical volume at a particular time point. The data snapshot can then be exported and used for restoring the volume data when required. The FusionStorage creates snapshots based on the distributed hash table (DHT) technology. Creating snapshots does not have any adverse impact on the volumes. For example, 24 MB space is required for constructing 2 million keys in the memory for a 2 TB hard disk. You can determine whether any snapshot has been created for the disk in a hash query operation. If a snapshot has been created, the hash query operation enables you to determine the storage location of the snapshot. The snapshot counts depend on the partition capacity. Because the partition capacity is not restricted, the FusionStorage theoretically supports infinite snapshots.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
20
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
3 System Composition of Huawei FusionCube for SAP HANA Appliance
Figure 3-11 High-performance snapshots provided by the FusionStorage
3.4.2 High-Performance Linked Cloning The FusionStorage provides the linked clone mechanism for incremental snapshots and creates multiple cloned volumes for a snapshot. The data in the cloned volumes is the same as that in the snapshot. Subsequent modification to a cloned volume does not affect the snapshot or other cloned volumes. The FusionStorage supports linked clone rate of 1:256 and improves utilization of storage space. The FusionStorage supports batch volume creation for VMs and can create hundreds of volumes in seconds. A cloned volume has all the functions of a common volume. You can create snapshots for a cloned volume, use the snapshot to restore the data in the cloned volume, and clone the data in the cloned volume.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
21
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
3 System Composition of Huawei FusionCube for SAP HANA Appliance
Figure 3-12 Linked cloning that the FusionStorage supports Base volume A
Cloned volume B/base volume B
Snapshot C1
Clone volume A Snapshot C2 Cloned volume B1
Cloned volume B2 Cloned volume C
3.4.3 Automatic Thin Provisioning FusionStorage provides the thin provisioning function, which allows users to use much more physical storage resources than real physical storage resources. This function remarkably improves storage utilization. FusionStorage uses the distributed hash technology to support automatic thinning provisioning, without pre-allocating space. Thin provisioning effectively prevents performance deterioration due to IP SAN capacity expansion. Figure 3-13 FusionStorage thin provisioning
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
22
Huawei Solutions for SAP HANA White Paper (Scale Out Solution - Distributed Storage)
3 System Composition of Huawei FusionCube for SAP HANA Appliance
3.5 Characteristics of Huawei FusionCube for SAP HANA Appliance
High performance: −
−
−
FusionStorage supports distributed parallel utilization for memory in the storage controller and for storage disks in PCIE-SSDs, thereby providing higher IOPS and I/O bandwidth. Logs in HANA appliance must be synchronized into the persistent storage media, which is the key factor that affects the system performance. In Huawei FusionCube for SAP HANA, logs are written to the NVDIMM buffer (non-volatile dual in-line memory module developed by Huawei and having the write and read performance as memory) of each storage server, which achieves better performance. Huawei converged network architecture and IB high-speed storage network remove the network bottleneck issue and improve performance for inter-node communication.
High reliability: −
FusionStorage in Huawei FusionCube solution for SAP HANA appliance ensures high data reliability. FusionStorage supports two or three data copies. If one or two storage nodes are faulty, upper-layer applications will not be affected.
−
In Huawei scale out solution, the computing system is separated from the storage system. Therefore, faults in the storage system exert no adverse impact on the computing system.
−
In Huawei scale out solutions, the redundancy strategy is designed for networks, ports, switches, and subracks. Therefore, no single node fault occurs.
Low TCO: − Huawei solution uses the converged architecture, which achieves higher device density, less space occupation, and less power consumption. − −
Some core components (such as PCIE-SSD) are developed by Huawei himself, which reduces costs, power consumption, and spaces. In Huawei FusionCube for SAP HANA, servers, storage devices, and switches are provided by Huawei. Therefore, all hardware devices can be managed in a centralized and unified manner, reducing maintenance costs.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
23
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
4
Key Solutions of Huawei FusionCube for SAP HANA Appliance 4.1 Cluster Solution Huawei FusionCube solution for SAP HANA appliance adopts the cluster architecture with computing physically separated from the storage. The following figure shows the logical cluster architecture of Huawei FusionCube for SAP HANA. Figure 4-1 Cluster logical architecture of Huawei FusionCube for SAP HANA
The system includes:
Computing cluster: responsible for installing HANA database and in-memory computing
Storage cluster: responsible for persistent storage for Log volume and Data volume Shared storage: provides storage space for Shared volume
Unified management: responsible for unified monitoring and management to computing/storage/network.
Network:
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
24
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
IB network: used for read/write of Log&Data volume
10GE network : used for inter-node communication, read/write for shared volume, system management and business connection.
Characteristics of Huawei cluster solution:
Separation of the computing from the storage: Computing part of the system constitutes a cluster while the storage part of system constitutes another cluster. The two clusters communicate using the shared file system or distributed block storage. Volumes are separated logically: Log volumes , Data volumes and shared volume are logically separated for deployment.
Volumes can be separated or combined physically: Log volumes , Data volumes and Shared volume can be deployed either in the same physical storage pool or in the different physical storage pools.
Agile storage media: The SSD or SAS can be used as the primary storage.
Advantages of the separated architecture:
Fully using key resources of the E7 blade: No much storage software is running on the computing node. Therefore, expensive E7 CPU resources can be fully used by the SAP HANA database to ensure its performance. Good fault separation ability: Computing nodes and storage nodes do not affect each other. For example, if a storage node is faulty, data must be re-established, occupying large CPU usages. However, the re-establishment is not performed on the SAP HANA computing blade. Therefore, the SAP HANA database performance is not affected. High expansion flexibility: Huawei adopts blade servers to establish the storage pool. With the enhancement of the computing capacity due to CPU updating, larger memory unit capacity, the expansion can be implemented by simply adding storage blade servers. In addition, original storage blade servers can still be used regardless of how greatly HANA nodes in computing change.
4.2 Data Persistence Solution 4.2.1 Working Principle During the normal running of the SAP HANA database, all data is stored in the memory. To ensure that the system can still be functional after the system is powered off, persistent data storage is required. The following figure shows the working principle of the data persistence solution.
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
25
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
Figure 4-2 Working principle of the data persistence solution
The following two types of data should be persistent:
Log: This is the information about data changes, that is, the redo log information. The logs are written into the persistent storage media after the transaction is complete. According to SAP requirements, the SSD or SAS disk is used as the persistent storage media. If the Log volume is full, logs will be backed up into the Data volume. Then, new data is written to the Log volume with an overwrite mode.
Data: Includes SQL data, undo log information, and appendix SAP HANA information (such as modeling data). According to SAP requirements, HDD is recommended as the storage media. The SSD can also be used as the storage media, at that time the system with SSD endures less service interruption time in the case of system restart, backup, and restoration.
Writing operation:
Log: Writing logs is a synchronous operation. Specifically, after a transaction is complete, the SAP HANA database returns success to the user only after data is successfully written into the Log volume. Therefore, the write operation affects the performance of the SAP HANA database.
Data: Writing data is an asynchronous operation, which is performed at the interval of 5 minutes by default. The interval is configurable. In savepoint mode with increment used, the written data is data changes and undo logs. Whether to perform the write operation is determined by memory changes. The SAP company has requirements on writing bandwidth under the file system layer. For example, the minimum requirement on 1 MB block is 200 Mbit/s. When the lower layer hardware is faulty and the write operation fails, the SAP HANA software generates an alarm.
The Log write operation is a synchronous operation, which affects the performance of the SAP HANA database. To ensure the high performance of the SAP HANA database, It is needed that performing the write operation as quickly as possible.
4.2.2 Media for Storing Persistent Data Physical media
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
26
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
Huawei Westmere-based 512 GB scale out solution adopts SSDs as the primary persistence storage for Log and Data. Huawei Ivybridge-based scale out solution (512 GB and 1 TB) adopts the SSDs as the default storage for Log volume and SAS disk for Data volume. You can also use the SAS disk for storing Log information. The SAS disk is used as the persistence media for storing Data information by default.
SSD Huawei FusionCube for SAP HANA adopts ES3000 PCI-E SSDs developed by Huawei. Huawei ES3000 PCI-E SSD is the PCI-E SSD that has the fastest storage speed in industries and has been evaluated by StorageReview, an international authoritative evaluation organization. Compared with Fusion-IO, the ES3000 OLTP has 16% higher performance and only 86% of the average response time.
PCI-E interface, which enables the appliance to communicate with the CPU. Provides bandwidth performance that is 1.5 to 2.2 times and IOPS performance that is 3 to 7 times as high as that of peer competitors.
High IOPS: Supports a maximum of 770000 IOPS for 4 KB random reads and 630000 IOPS for 4 KB random writes.
High bandwidth: Provides eight PCI-E 2.0 interfaces, which improves the bandwidth up to 3200 Mbit/s.
Low access delay: 8 µs
Provides large capacity (2.4 TB) for a flash card and long lifecycle (19.1 PB).
Provides a two-dimensional error checking and correcting (ECC) mechanism, improving reliability by up to six orders of magnitudes.
Supports the Read Retry technology, which increases the lifecycle of a Flash card by more than 3 times.
Uses a built-in data scrubbing engine to periodically check stored data, preventing data errors.
Uses the dynamic RAID algorithm to achieve resource sharing between channels to prevent adverse impact due to chip failures.
Provides the power failure protection mechanism, thereby ensuring data integrity. Supports both dynamic and static wear leveling algorithms.
SAS Disk
900/1200G/1800G
10K SAS disk
4.2.3 Cache Acceleration Mechanism Huawei FusionCube for SAP HANA uses self-developed ES3000 v2 SSD as the cache, which further improves the performance of synchronous writing of logs and achieves shorter writing delay in the SAP HANA certification.
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
27
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
4.3 Service Continuity Assurance 4.3.1 HA Basic Concept The service continuity is ensured through the implementation of high reliability (HA). HA is a series of technology, engineering practice, and design principle that can ensure service continuity. The HA is implemented by avoiding single point faults (fault tolerance) and improving recovery speeds after faults occur (fault recovery). Huawei FusionCube for SAP HANA adopts the following HA solutions:
Avoiding single point faults: The system adopts the redundancy design to avoid single point faults. − −
−
Hardware redundancy: The E9000 subrack adopts the redundancy design for dual-switching plane, fans, power supply, and management module. Network redundancy: −
For inter-subrack: Full redundant connection is used between each subrack and aggregation subrack.
−
For inter-board: Each board provides two 10GE NICs with 4 network ports for each. Two network ports on different NICs constitute a bond to ensure the redundancy for network communication.
Data center redundancy: With customer requirements, the production site and DR sites are deployed. When production site is faulty, DR site takes over services.
Fault recovery capability: −
−
−
−
Service auto restart: If some service processes of the HANA, for example, NameServer or IndexServer software is faulty, the HANA monitoring service starts these processes to ensure the service continuity. Host auto-failover: Multiple worker nodes and standby nodes are deployed in the same local system (Production site or DR site). When a worker node is faulty, its standby node takes over services. Backup: Backs up data in the SAP HANA database to an outer storage. If the system breaks down or data is damaged, reload the backup data to the memory to recovery services. Disaster (DR) replication: Copies data of primary nodes to one redundant site. When all primary nodes are powered off or cannot provide services due to fire and earthquake, redundant nodes take over services. The DR replication can be implemented using the following two modes: − Storage replication: The DR replication is implemented in the storage layer. The SAP HANA database cannot sense it. (This mode is provided only for SAN storage.) −
System replication: The DR replication is implemented by the SAP HANA software.
The HA solution to be adopted is determined based on customer requirements and implementation costs. Comparison for HA solutions is as follows: Solution
Advantage
Disadvantage
Host auto-failover
Issue 1.0 (2014-01-26)
Performs automatic detection and switchover.
Standby nodes must have access the shared storage.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
28
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
Solution
Advantage
Disadvantage
Backup
Used for redundancy recovery.
Features lowest costs and simplest implementation. Supporting the recovery at specified time.
The RPO varies from minutes to hours depending on backup frequency.
Standby system deployment may cost hours or multiple days.
The cold start mode is adopted. RTO requires several hours.
The column data loading may cost hours to enter the optimal performance status.
During the service takeover, it takes hours to clear software and data in the original system. Then, you can restart the HANA system.
The cold start mode is adopted. RTO requires some hours. The system enters the optimal performance status after multiple hours.
Storage replication
Can also be used for cloning or replicating systems.
Used for redundancy recovery. In synchronous replication, RPO is 0. In asynchronous replication, RPO takes only seconds. Redundancy systems can be used for other functions when they are not used.
System replication
Used for redundancy recovery.
In synchronous replication, RPO is 0. In asynchronous replication, RPO takes only seconds.
In pre-loading mode, RTO is only in minute level.
The system reaches optimal performance immediately after the switchover. Compatible with solutions provided by all hardware vendors.
Issue 1.0 (2014-01-26)
The network-based storage system and high bandwidth connection between sites are required. The synchronous replication can be implemented only within 100 km. Replication consumes more bandwidth than the system replication.
The system cannot be protected when the storage breaks down.
This solution depends on support of different hardware vendors.
In pre-loading mode, live and dedicated redundancy offices are required.
Effective connections between offices are required. After the switchover, automatic reconnection solution on the client is required. The DNS or virtual IP solution is used in general.
Supports the single node
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
29
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
Solution
Advantage
Disadvantage
system with local storage.
In non-preloading mode, redundant offices can be used for other functions when they are not used.
The following table lists RPO and RTO of HA solutions. Scenario
HA solution
Implementation Cost
RPO
RTO
Local fault recovery
Services automatic restart.
0
0
Medium
Node automatic switchover
Medium
0
Medium
Data backup and recovery
Low
0
High
Storage replication
Medium
Near to 0.
Medium
Redundancy recovery
System replication
High
0
Switchover costs 1 to 3 minutes.
10 to 20 minutes for hot start
Some hours for cold start.
Recovery of the full performance costs hours.
Low
2 to 5 minutes for preloading
10 to 20 minutes for non-preloading costs. Full performance can be immediately recovered.
4.3.2 Host Auto-Failover Huawei FusionCube for SAP HANA provides the failover capability of nodes in the same data center. If a node is faulty in a cluster, the system can automatically detect the fault and
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
30
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
trigger a failover between the faulty node and standby node, thereby ensuring service continuity. Huawei FusionCube for SAP HANA provides the host auto-failover capability to eliminate single point of failure risks. The system supports a maximum of 16 HANA nodes, including one standby node. Multiple standby nodes can be deployed, but the number of standby nodes cannot exceed the number of worker nodes, Therefore, the system can ensure service continuity if multiple primary nodes are faulty. The system that provides the host auto-failover function has the following advantages:
Ensuring service continuity: For a system that does not provide the auto-failover capability, if a running node is faulty, services provided by this node are interrupted. For Huawei FusionCube for SAP HANA, if a running node is faulty, its standby node takes over services immediately after loading data.
Reducing interruption time: Before upgrading the system, switch over services from the active nodes to the standby nodes to ensure the service continuity even during upgrade.
Working principle:
In the following figure, HANA node 2 is faulty. The master node responsible for cluster management detects the fault and instructs standby node to take over services running on node 2. Attach node 2's volumes to the standby node m, loads data from node 2's volume, and then provides services. During the switchover, the SAP HANA database software invokes storage interface scripts, which implement the detaching and attaching of Log and Data volumes, so that the volume attached to node 2 switches to the standby node.
After the fault of node 2 is rectified, Huawei provides the I/O fencing function at the storage layer to prevent node 2 from writing data into the original volume.
Figure 4-3 Working principle of the node switchover
The switchover occurs in one of the following scenarios:
Network fault: Heartbeat loses between HANA nodes (faults may occur on NICs, cables, and ports).
Storage fault:
Issue 1.0 (2014-01-26)
−
The Log, Data, or Share volume of the node is unreachable.
−
The storage software agent on the HANA node is faulty.
OS-level fault: The node OS is faulty. For example, the file system is faulty.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
31
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
Database software fault: Main HANA database software processes, such as NameServer and IndexServer, are faulty and cannot be restarted by the monitoring process Daemon.
4.3.3 Backups (Data Backup and Recovery Solution) Huawei FusionCube solution for SAP HANA appliance provides two types of data backup solution.
4.3.3.1 Shared NAS Backups Principle The customers use an independent NAS system or a centered NAS system to provide shared storage space for data backups of SAP HANA database system. The principle is as follows： Figure 4-4 Working principle of NAS backups
Backups of Log and Data volumes are all written into this shared NAS； Log backup is auto by default. Data backup can be triggered by operations in SAP HANA studio, SQL commands or DBA scripts; data backup can be triggered once or periodically
NAS device selection Customer can buy N8500/N9000 NAS, OceanStor T-series unified storage from Huawei or any NAS device from other suppliers for shared backup NAS. By default, we recommend N8500 NAS from Huawei. In this time, there is as following figure about shared NAS network deployment.
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
32
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
Figure 4-5 NAS backup network
In huawei network deployment for SAP HANA appliance, each HANA node has 8*10GE ports and consists of 4 bonds. Backup panel and business panel have its independent bond. In this situation, backup operation has no negative effect on business network bandwidth, so we can refer to this backup method as LAN-Free backup.
4.3.3.2 3rd Party Backup Software Backups Principle 3rd party backup software which has passed the verification of SAP can be used for implementing SAP HANA backup. Its principle is as follows： In the HANA computing node 3rd party backup agent is installed Backup agent reads data from persistent data area and send them to outer backup server Backup server sends data to backup destination. Figure 4-6 3rd party backup software
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
33
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
The list of SAP-certified 3rd party backup software is as follows:
According to SAP Note 0001730932, although these software have passed certification of SAP, when they are used in Huawei FusionCube SAP HANA Appliance they will still need to pass a compatible test with Huawei infrastructure, otherwise they can’t be delivered to customers.
A example of 3rd party backup software in Huawei SAP HANA Appliance If Commvault backup software is used and backup destination is outer tapes, the whole system including HANA system and backup system will be as follows: Figure 4-7 A FusionCube SAP HANA appliance backup solution with Commvault 3 rd party backup software
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
34
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
HANA appliance is located in E9000 chassis. There are many HANA computing nodes and storage nodes in E9000 chassis. In each computing node media server software and backup agent software from Commvault will be installed. Customer need buy some backup devices except that E9000 HANA appliance. o A server as backup management server, backup policy can be set in this server. o A management switch serving as the switching function between backup management server and HANA appliance o A set of tapes as target backup system. o One or two backup switch serving as the storage switching function between HANA appliance and target tapes. In this example, backup tapes only can provide FC interface, so backup switch should be FC switch, and in E9000 chassis CX311 switch module will be setup to provide FC switch capability.
4.3.4 Disaster Recovery Working Principle The disaster recovery solution based on system replication provided by the SAP HANA database software ensures RPO = 0 and RTO in minute level. Customers are recommended to use this disaster recovery solution for Huawei FusionCube for SAP HANA. You can choose one of the following three data replication modes as shown in the following figure:
Synchronous: Data is written into the memory of the primary node, the primary site sends the data to the secondary node, the data is written into the memory of the secondary node, and then the primary site receives a success response.
Synchronous in-memory: Data is successfully written into the memory of the secondary node. Asynchronous: The primary node successfully sends Log data to the secondary node and does not need to wait for success responses.
Figure 4-8 System replication in synchronous mode
Disaster recovery effects for synchronous and asynchronous modes
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
35
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
Synchronous
Distance
RPO
RTO
< 100 km
0
2 to 5 minutes for preloading
10 to 20 minutes for non-preloading
Asynchronous No special restrictions
Second level
Hour level
Before using the synchronous replication, ensure that configurations of the primary and secondary sites must be the same.
Client Application connection Mode If the disaster recovery solution based on system replication is used, after the HANA database switchovers, the client must use one of the following methods to reconnect to the secondary HANA database for service continuity.
Method 1: IP redirection Application scenario
Both the primary and secondary sites are in layer 2 network. This applies to the scenario where the primary and secondary sites are in the same city and the synchronous replication is used by default.
Figure 4-9 IP redirection
Working principle: 1.
The client accesses the HANA system using the logical host name. The client first accesses the DNS for host name resolution.
2.
The DNS stores a logical host name and a logical IP address (virtual IP address) for each pair of primary/secondary nodes. The DNS replies with a virtual IP address corresponding to the logical host name.
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
36
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
3.
The client accesses a certain node of the HANA system using the virtual IP address through the layer 2 switch.
4.
The layer 2 switch uses a script or configuration to map the virtual IP address to a node port of the primary site. If the node in the primary site is faulty, the virtual IP address is mapped to a node port of the secondary site. (Each node has such a mapping.)
5.
In fault scenarios, the switch redirects the client to the secondary site.
Advantages:
No configurations are required on the client.
The recovery is fast. The method for using the SQL and HTTP clients are the same.
Disadvantages:
The primary and secondary nodes must be on the same layer 2 network.
Method 2: DNS resolution Application scenario Primary and secondary sites are deployed on different layer 3 networks. This applies to the scenario where the primary and secondary sites are in different cities and the asynchronous replication is used by default. Figure 4-10 DNS resolution
Working principle: 1. 2. 3.
The client accesses the HANA system using the logical host name. The client first accesses the DNS for host name resolution. The DNS replies with an IP address of a node in the primary site. The client uses this IP address to access the HANA system. If the client finds that the primary site is faulty, the client sends a redirection request to the DNS. The DNS replies with the corresponding node's IP address in the secondary site. The client uses this new IP address to access the HANA system.
Advantages:
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
37
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
No configurations are required on the client.
Active and secondary sites can be on different network, that is, separated by routers.
Disadvantages:
DNS mapping must be modified, because mapping principles may be different for different DNS vendors.
Switchover on the client lasts for a long time, because the update of the DNS cache all over the network requires some time and even some hours.
4.4 Networking 4.4.1 FusionCube E9000 4.4.1.1 Logical Network Design Huawei FusionCube SAP HANA cluster appliance has redundant network design. The system can provide up to 5 physical communication panels to assure high performance and high availability. Figure 4-11 Logical Network design of Huawei SAP HANA Appliance
Ethernet switch: Several pairs of 10GE switch modules in all E9000 chassis implement Ethernet switch.
Issue 1.0 (2014-01-26)
o
Appliance management and HANA business share a pair of 10GE
o
HANA inter-node communication uses a pair of 10GE independently
o
Backup uses a pair of 10GE independently
o
Disaster recovery uses a pair of 10GE independently
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
38
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
Infiniband(IB) switch: Several pairs of IB switch modules in all E9000 chassis implement infiniband switch o
Persistent storage uses a pair of 56G infiniband independently between HANA node and storage node.
Outer interface: The system has up to 6*10GE + 16*8Gb FC ports for outer connection through 10GE switch module in the E9000 chassis.
4.4.1.2 Network Interconnection The following factors affect the networking solution:
Node bandwidth − −
The 10GE transmission between two HANA nodes must be ensured. Network bandwidth (10GE is recommended) between the computing and storage nodes must be ensured.
Power consumption: No greater than 9 kw for each subrack
Easy expansion: Two HANA nodes can be expanded to 16 HANA nodes. The modification workload of system connections is light.
The following figure shows the networking solution for a scale out system with Sixteen 1 TB HANA nodes (CH223 used as the Log volume, and CH222 used as the Data and Share volumes). Figure 4-12 Networking solution of Huawei FusionCube E9000 for SAP HANA appliance
Maximum cabinets: The system consists of five subracks and two cabinets.
Hybrid deployment: The hybrid deployment is adopted by default to reduce power consumption. Specifically, for the first four subracks (OSCA1, OSCA2, OSCA3, and OSCA4), a maximum of four HANA nodes and four storage nodes are configured for
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
39
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
each subrack. For the fifth subrack (OSCA5), eight nodes are configured as storage nodes.
Inter-subrack connection using the IB network (lines in blue): −
Inter-subrack interconnection: The first two subracks are used as the connection subracks. The last three subracks connect to the first two subracks using the IB network. Each IB switching board in each subrack provides eight IB cables for the connection to the first two subracks.
−
Board interconnection: Each HANA or storage node has one IB NIC that provides two IB network ports. Eight nodes in each subrack must provide 16 IB cables to ensure the IB bandwidth at a maximum degree. As shown in the preceding figure, each subrack provides at least 16 IB cables, which fully ensures the bandwidth for IB network interconnection.
Inter-subrack interconnection for 10GE network (lines in black): −
As the central switching subrack for the 10GE network, the first subrack connects to all other subracks over the Ethernet.
−
Each of the first four subracks provides a maximum of four HANA nodes and eight 10GE connections on the primary/secondary plane to the first subrack, ensuring the sufficient 10GE connection for each HANA node.
−
The fifth subrack does not provide HANA nodes. Therefore, only two 10GE connections are required for the connection to the first subrack. These connection are used in the management plane of the FusionCube integration architecture.
Crossover cables are used to connect subracks so that the system is still functional when a cross fault occurs on the inter-subrack network.
System outlet cables: The first subrack is used as the basic subrack and provides (3 + 3) 10GE connections. One is used as the service cable, one as the data replication cable (for data replication import, export, backup, and DR), and one as the management cable (for the management of HANA studio, FusionCube, and HANA client).
4.4.1.3 Port and Bond Solution Port design: Each HANA node provides three NICs.
There are two 10GE NICs, each of which provides four ports. − Ports 1 and 6 are used for internal communication of the FusionCube infrastructure. −
Ports 2 and 5 are used for internal communication of the HANA nodes.
−
Ports 3 and 8 and ports 4 and 7 provide external IP addresses. Which two ports are used is determined by the customer. You can divide multiple WLANs on the two bonds to provide multiple external IP addresses.
One IB NIC, which provides two 56 Gbit/s IB network ports to constitute a bond port. Together with the storage node, a storage network is formed.
Each storage node provides two NICs.
One IB NIC, which provides two 56 Gbit/s IB network ports to constitute a bond port. Together with the HANA node, a storage network is formed. The network provides storage channels between storage nodes.
One 10GE NIC, which provides two 10GE network ports for the internal communication of the FusionCube infrastructure.
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
40
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
Figure 4-13 Port solution of Huawei FusionCube for SAP HANA
Bond design Each HANA node provides eight 10GE ports. Two ports on different switching planes of different physical cards constitute a bond port, preventing single port faults and eliminating cross-connection faults.
Ports 1 and 2, 3 and 4, 5 and 6, and 7 and 8 belong to the same network chip for each pair, which respectively connect to the primary and secondary switching boards.
Ports 1, 2, 3, and 4 belong to one physical NIC, while ports 5, 6, 7, and 8 belong to another physical NIC. Ports 1 and 6 constitute bond 0, 2 and 5 constitute bond 1, 3 and 8 constitute bond 2, and 4 and 7 constitute bond 3.
Figure 4-14 Bond solution of FusionCube for SAP HANA
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
41
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
4.4.1.4 Networking Feature Huawei FusionCube solution for SAP HANA appliance features the following:
External switch device are not required in a scale out system with less than 16 nodes.
All networks are deployed in redundancy mode to prevent single-point network failure risks and cross-connection faults, thereby ensuring high service availability.
The communication between HANA nodes uses an exclusive 10GE network plane to ensure high service bandwidth. HANA nodes communicate with storage nodes using the 56 Gbit/s IB network to ensure high storage bandwidth and smooth communication.
4.4.2 Fusioncube RH8100 v3 4.4.2.1 Logical Network Design Huawei FusionCube SAP HANA cluster appliance has redundant network design. The system can provide up to 5 physical communication panels to assure high performance and high availability. Figure 4-15 Logical Network design of Huawei SAP HANA Appliance
Ethernet switch: Several pairs of 10GE switch modules in all E9000 chassis implement Ethernet switch.
Issue 1.0 (2014-01-26)
o
Appliance management and HANA business share a pair of 10GE
o
HANA inter-node communication uses a pair of 10GE independently
o
Backup uses a pair of 10GE independently
o
Disaster recovery uses a pair of 10GE independently
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
42
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
Infiniband(IB) switch: Several pairs of IB switch modules in all E9000 chassis implement infiniband switch o
Persistent storage uses a pair of 56G infiniband independently between HANA node and storage node.
4.4.2.2 Network Interconnection
Every RH8100v3 have 8*10GE portal and 2*IB portal
Figure 4-16 Networking solution of Huawei FusionCube E9000 for SAP HANA appliance
4.5 Capacity Expansion and Reduction 4.5.1 Capacity Expansion Application scenario When service data of the SAP HANA database increases and the memory capacity is insufficient, users need to buy new nodes for capacity expansion.
Capacity expansion method
Issue 1.0 (2014-01-26)
For Westmere-based systems, perform the capacity expansion using the scale-out mode.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
43
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
For Ivybridge-based systems, perform the capacity expansion using the scale-up or scale-out mode. −
If 512 GB nodes are in use, you can use the scale-up mode to upgrade the capacity of each HANA node to 1 TB. In this situation, you must replace all 16 GB memories with 32 GB memories.
−
If 1TB nodes are in use, you can only use the scale-out mode for capacity expansion.
Expansion unit The basic principle is that the expansion is performed by adding HANA nodes. Each expansion has different requirements depending on whether the expansion is performed with another different subracks, cabinets, or storage pools.
Basic cabinet: The first cabinet is the basic cabinet. Basic subrack: The first subrack that contains management nodes is the basic subrack. Extension subrack: Subracks except the basic subrack are extension subracks. Startup unit: two SAP HANA nodes (including two computing nodes and three storage nodes) for Huawei FusionCube for SAP HANA Expansion unit: one SAP HANA node. When adding an SAP HANA node, add a computing node. The number of storage nodes to be added may be 0, 1, or 3, which depends on whether the expansion is performed with another different subracks, cabinets, or FusionStorage resource pools. For details about the number of boards, see section 6.3.3 "Configuration by Memory".
Impact on services Data is migrated during capacity expansion. Therefore, before the expansion, services must be stopped.
4.5.2 Capacity Reduction Application scenario The user has deployed multiple SAP HANA systems and wants to adjust configurations in different SAP HANA clusters. Retails or the SAP Company needs to uninstall one or more nodes from the entire SAP HANA cluster for POC test in customer sites.
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
44
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
M ake a proper data planning before reducing the capacity, ensuring that the capacity of remained disks is sufficient for storing all data.
Stop services before starting the capacity reduction.
Perform the capacity reduction on computing nodes first and then on storage nodes.
4.6 Other Solutions 4.6.1 Unified Management Solution
Huawei FusionManager manages infrastructure (including computing, storage, and network) of Huawei FusionCube for SAP HANA. The unified management reduces management difficulties and system maintenance TCO.
The client studio of the SAP HANA database provides monitoring and management services for the HANA database system.
FusionManager will provide the northbound interface in the future, which enables FusionManager to report infrastructure alarms to the unified monitoring platform Solution Manager provided by the SAP. This achieves unified monitoring and management for Huawei FusionCube for SAP HANA.
4.6.2 Multiple SID deployment Multiple SID HANA databases can be installed in a set of HANA appliance. According to the requirement from SAP, multi-SID system can only be used in DEV/QA scenario, not in production scenario. Required resource sum from multiple SID HANA database systems can’t exceed the real physical resource.
4.6.3 Cluster split and consolidation
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
45
iSite BTS3001C-116
4 Key Solutions of Huawei FusionCube for SAP HANA Appliance
Customer may want split one HANA cluster appliance of 8 nodes into two cluster appliances with each having 4 nodes in some scenario. Customer may also want consolidate one appliance of 2 nodes and the other appliance of 6 nodes into one cluster appliance of 8 nodes in other scenario. Huawei can provide this capability of HANA cluster split and consolidation, but there are some limits as following: Extra service charge from Huawei will be needed for implementing cluster split and consolidation. It is not any condition in which split and consolidation can be implemented. on real condition.
It depends
During implementing cluster appliance split/consolidation, in general business will be down for a while. If customer wants to no downtime of business, it depends on real condition.
4.6.4 Virtualization Solution The virtualization solution is pre-configured in the management nodes of Huawei FusionCube for SAP HANA. If the SAP Company agrees to run the SAP HANA database on a virtualized platform to provide cloud services, Huawei will deploy the HANA database on Huawei virtualization platform.
4.6.5 Compatibility Hardware:
The Westmere board with 4 CPUs is used on the E9000. The backward compatibility is not considered, that is, Westmere boards are still used for expansion for offices using Westmere boards. Ivybridge boards are the mainstream boards for commercial delivery.
OS:
The authenticated computing node uses Suse Linux 11.2 or 11.3 for SAP or the latest version according to SAP requirements. The compatibility of HANA data to the OS is ensured by the SAP Company. Huawei is responsible for updating OS drives. The authenticated storage node uses redhat 6.2. FusionCube infrastructure ensures the software compatibility.
FusionStorage software:
The FusionStorage protects data from being lost during software upgrade. There are no special requirements for SAP HANA appliance.
SAP HANA database software:
Issue 1.0 (2014-01-26)
The SAP company ensures the compatibility. The original database data is used after the software upgrade.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
46
iSite BTS3001C-116 5 Specifications of FusionCube for SAP HANA Appliance
5
Specifications of FusionCube for SAP HANA Appliance 5.1 Delivery Scope of FusionCube for SAP HANA Appliance Figure 5-1 shows the delivery scope of FusionCube for SAP HANA. Figure 5-1 Delivery scope of FusionCube for SAP HANA
Items highlighted in red, including hardware, software, installation, and maintenance, are provided by Huawei.
Hardware: indicates FusionCube infrastructures, including cabinets, power distribution modules, subracks, blades, and switch modules.
Software: includes SUSE OSs (3-year warranty) and FusionStorage. Installation: includes installation of hardware, OSs, and HANA databases.
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
47
iSite BTS3001C-116 5 Specifications of FusionCube for SAP HANA Appliance
Maintenance: includes all one-stop hardware and software warranty services provided by Huawei.
5.2 Specifications of FusionCube for SAP HANA Appliance Table 5-1 lists the specifications of FusionCube for SAP HANA. Table 5-1 Specifications of FusionCube for SAP HANA（*1 ） CPU Platform Specification s Name
Cluster scale
Item
Number of HANA nodes
Westmere EX Platform (E7V1)
Ivybridge EX Platform (E7V2)
512 GB Node
512 GB Node
High-
High-
Performance Edition
Performance Edition
Normally 1 to 15 (16 nodes are also supported, but which require additional processing during installation and delivery)
1 TB Node( *2) Standard Edition
High-
1 to 16(*3)
1 to 16
1 to 16
1 to 16
Performance Edition
Standard Edition
Default: 2 nodes with one as the active node and the other as the standby node Computing node (CH242)
CPU
4 x E7 4870 (2.4 GHz)
4 x E7 4890 (2.8 GHz) or 4880 (2.5 GHz)
4 x E7 4890 (2.8 GHz) or 4880 (2.5 GHz)
4 x E7 4890 (2.8 GHz) or 4880 (2.5 GHz)
4 x E7 4890 (2.8 GHz) or 4880 (2.5 GHz)
Memory
512 GB (32 x 16 GB)
512 GB (32 x 16 GB)
512 GB (32 x 16 GB)
1 TB (32 x 32 GB)
1 TB (32 x 32 GB)
System disk
2 x 600 GB 10K SAS
2 x 600 GB 10K SAS
2 x 600 GB 10K SAS
2 x 600 GB 10K SAS
2 x 600 GB 10K SAS
Network port
6 x 10GE+2 x IB
8 x 10GE+2 x IB
8x 10GE+2 x IB
8 x 10GE+2 x IB
8x 10GE+2 x IB
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
48
iSite BTS3001C-116 5 Specifications of FusionCube for SAP HANA Appliance
CPU Platform Specification s Name
Storage node (CH223)
Storage node (CH222)
Management node
Westmere EX Platform (E7V1)
Ivybridge EX Platform (E7V2)
512 GB Node
512 GB Node
High-
High-
Performance Edition
Performance Edition
OS
SUSE Linux 11.3 for SAP
SUSE Linux 11.3 for SAP
CPU
2 x E5 2640
2 x E5 2640
2 x E5 2640
Memory
128 GB (16 x 8 GB)
128 GB (16 x 8 GB)
128 GB (16 x 8 GB)
System disk
2 x 600 GB 10K SAS
2 x 600 GB 10K SAS
2 x 600 GB 10K SAS
Data disk
2 x 2.4 TB SSD
2 x 2.4 TB SSD
2 x 2.4 TB SSD
Cache
4 GB NVDIMM
4 GB NVDIMM
4 GB NVDIMM
Network port
2 x 10GE+2 x IB
2 x 10GE+2 x IB
2 x 10GE+2 x IB
OS
Red Hat 6.2
Red Hat 6.2
Red Hat 6.2
Item
1 TB Node( *2) Standard Edition
High-
SUSE Linux 11.3 for SAP
SUSE Linux 11.3 for SAP
Performance Edition
Standard Edition
SUSE Linux 11.3 for SAP
CPU
2 x E5 2640
2 x E5 2640
2 x E5 2640
2 x E5 2640
Memory
128 GB (16 x 8 GB)
128 GB (16 x 8 GB)
128 GB (16 x 8 GB)
128 GB (16 x 8 GB)
System disk
2 x 900 GB 10K SAS
2 x 900 GB 10K SAS
2 x 900 GB 10K SAS
2 x 900 GB 10K SAS
Data disk
12 x 900 GB 10K SAS
12 x 900 GB 10K SAS
12 x 900 GB 10K SAS
12 x 900 GB 10K SAS
Cache
4 GB NVDIMM
4 GB NVDIMM
4 GB NVDIMM
4 GB NVDIMM
Network port
2 x 10GE+2 x IB
2x 10GE+2 x IB
2 x 10GE+2 x IB
2x 10GE+2 x IB
OS
Red Hat 6.2
Red Hat 6.2
Red Hat 6.2
Red Hat 6.2
2 x E5 2640
2 x E5 2640
2 x E5 2640
2 x E5 2640
CPU
Issue 1.0 (2014-01-26)
2 x E5 2640
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
49
iSite BTS3001C-116 5 Specifications of FusionCube for SAP HANA Appliance
CPU Platform Specification s Name
Cabinet density
Power consumption
Westmere EX Platform (E7V1)
Ivybridge EX Platform (E7V2)
512 GB Node
512 GB Node
High-
High-
Performance Edition
Performance Edition
Memory
128 GB (16 x 8 GB)
System disk
Item
1 TB Node( *2) Standard Edition
High-
128 GB (16 x 8 GB)
128 GB (16 x 8 GB)
128 GB (16 x 8 GB)
128 GB (16 x 8 GB)
2 x 900 GB 10K SAS
2 x 900 GB 10K SAS
2 x 900 GB 10K SAS
2 x 900 GB 10K SAS
2 x 900 GB 10K SAS
Network port
2 x 10GE
2 x 10GE
2 x 10GE
2 x 10GE
2 x 10GE
OS
SUSE Linux 11.2
SUSE Linux 11.2
SUSE Linux 11.2
SUSE Linux 11.2
SUSE Linux 11.2
Single-ca binet density
3 subracks, 11 HANA nodes (10 worker and one standby), and 5 TB available RAM
3 subracks, 14 HANA nodes, and 6.5 TB available RAM
3 subracks, 15 HANA nodes, and 7 TB available RAM
3 subracks, 10 HANA nodes, and 9 TB available RAM
3 subracks, 11 HANA nodes, and 10 TB available RAM
Full-confi guration density(16 nodes)
2 cabinets, 5 subracks, and 7 TB available RAM
2 cabinets, 4 subracks, and 7 TB available RAM
2 cabinets, 4 subracks, and 7 TB available RAM
2 cabinets, 4 subracks, and 15 TB available RAM
2 cabinets, 4 subracks, and 15 TB available RAM
Computin g subrack
6.5 KW (eight computing nodes in one subrack)
10 KW (eight computing nodes in one subrack)
10 KW (eight computing nodes in one subrack)
10 KW (eight computing nodes in one subrack)
10 KW (eight computing nodes in one subrack)
Storage subrack
2.5 KW (eight CH223 storage nodes in one subrack)
3.3 KW (eight CH222 storage nodes in one subrack)
3.3 KW (eight CH222 storage nodes in one subrack)
3.3 KW (eight CH222 storage nodes in one subrack)
3.3 KW (eight CH222 storage nodes in one subrack)
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
Performance Edition
Standard Edition
50
iSite BTS3001C-116 5 Specifications of FusionCube for SAP HANA Appliance
CPU Platform Specification s Name
Weight
Westmere EX Platform (E7V1)
Ivybridge EX Platform (E7V2)
512 GB Node
512 GB Node
High-
High-
Performance Edition
Performance Edition
Composit e subrack
3.7 KW (three computing nodes, three storage nodes, one management node, in one subrack)
Full configurat ion of one carbinet
Item
1 TB Node( *2) Standard Edition
High-
N/A
6.7 KW (four computing nodes, three storage nodes, one manageme nt node, in one subrack)
N/A
6.7 KW (four computing nodes, three storage nodes, one manageme nt node, in one subrack)
15 KW (three subracks)
21.9 KW (three subracks)
23.9 KW (three subracks)
19.2 KW (three subracks)
21.3 KW (three subracks)
Computin g subrack
220 kg
220 kg
220 kg
220 kg
220 kg
Storage subrack
220 kg
220 kg
220 kg
220 kg
220 kg
Composit e subrack
220 kg
220 kg
220 kg
220 kg
220 kg
Full configurat ion of one carbinet
820 kg (three subracks)
820 kg (three subracks)
820 kg (three subracks)
820 kg (three subracks)
820 kg (three subracks)
Performance Edition
Standard Edition
(*1): The information in above table is the result in the scenario of OCFS2 used for Shared volume. If NAS is used for Shared volume, there will be some difference: 1) need add a set of NAS device; 2) the number of storage node will be different; 3) Power consumption and weight will also be different; (*2): If CPU is not performance bottleneck in some scenario, 4CPU + 2T memory (32*64G) can be used to implement 2T size node; but before implementing this the project team in customer site need to get the admission from SAP at advance. (*3): 1) In the product scenario the minimum node number of SAP HANA cluster appliance is 2, this will assure the cluster system availability; 2) In some scenario, e.g. POC there can be only 1 HANA node in a HANA cluster; 3) Technically, Huawei SAP HANA appliance has a ability of more than 16 nodes but can be sold only after authenticated by SAP. If the customer poses intense requirements for the deployment of more than 16 nodes, Huawei will be engaged in the authentication work before conducting the POC tests.
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
51
iSite BTS3001C-116 5 Specifications of FusionCube for SAP HANA Appliance
Table 5-2 lists system reliability indicators. Table 5-2 System reliability indicators Item
Indicator (Tested Under No Service Load in the Eight-Node 512 GB System)
Node failover
RPO = 0 minute, RTO < 4 minutes Includes detecting a fault, triggering a failover, and loading data to the memory.
System restart upon a power failure
RPO = 0 minute, RTO < 20 minutes OS restart: < 15 minutes FusionStorage restart: < 3 minutes SAP HANA restart and system data loading: < 2 minutes
5.3 Performance of FusionCube for SAP HANA Appliance Backed by the distributed storage architecture, particular NVDIMM cache technology, and SSDs that serve as the main storage, Huawei FusionCube solution for SAP HANA appliance achieves outstanding performance several times higher than the base authentication data.
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
52
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
6
Configurations of FusionCube for SAP HANA Appliance 6.1 Configuration Principles of FusionCube for SAP HANA Appliance The configuration principles of FusionCube for SAP HANA appliance are as follows:
The system scale is determined by the amount of user data to be processed. The storage capacity can be adjusted based on whether an external backup center is deployed. By default, computing nodes and storage nodes are compositely deployed in one subrack. Alternatively, they can be deployed in independent subracks. Networking principles: −
The default networking plan does not involve an external switch. However, if the customer poses requirements for rapid capacity expansion, deploy external switches.
−
The first two subracks serve as the tandem subracks. The subracks of all storage blades connect to the tandem subracks using IB cables.
Node selection principles:
If the user requires every single node memory capacity smaller than 1TB, choice E9000 CH242v3 for DB nodes; If the user requires every single node memory capacity greater than 1 TB, choice RH8100v3 for DB nodes
The following table lists the recommended and default configurations and authentication of FusionCube for SAP HANA in different portfolios:
P indicates the recommended configurations in a certain scenario.
D indicates the default configurations in a certain scenario.
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
53
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
Table 6-1 Configurations and authentication of FusionCube for SAP HANA in different portfolios Node Portfolio
Authentication
Westmere
Ivybridge
512 GB
512 GB
High-
High-
Performance Edition
Performance Edition
Passed on October 2013
Passed on March 2014
1 TB Standard Edition
High-
Passed on March 2014
Passed on March 2014
Performance Edition
Standard Edition
Passed on March 2014
System capacity 1+1 scenario (for OLTP)
P, D (when the user requires only 512 GB capacity)
P, D (when the user requires 1 TB capacity)
1 TB available capacity
P, D
2 TB available capacity
P (preferential selection after expansion of the 512 GB all-SSD solution)
P
D
3 TB available capacity
P (preferential selection after expansion of the 512 GB all-SSD solution)
P
D
4 TB available capacity
P (preferential selection after expansion of the 512 GB high-performance solution)
D
5 TB available capacity
P
D
6 TB available capacity
P
D
7 TB available capacity
P
D
8 TB available capacity
P
D
9 TB available capacity
P
D
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
54
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
Node Portfolio
Westmere
Ivybridge
512 GB
512 GB
High-
High-
Performance Edition
Performance Edition
1 TB Standard Edition
HighPerformance Edition
Standard Edition
10 TB available capacity
P
P, D
11 TB available capacity
P
D
12 TB available capacity
P
D
13 TB available capacity
P
D
14 TB available capacity
P
D
15 TB available capacity
P
D
The customer requires a system with 4 TB available capacity. An example is provided as follows:
The system is newly built: − If the customer has no special requirements, use the Ivybridge-based 1 TB node system of the standard edition. −
Issue 1.0 (2014-01-26)
If the customer requires high performance, use the Ivybridge-based 1 TB node system of the high-performance edition. The customer expands an existing HANA system to the 4 TB system: −
If the existing system is the standard edition using the 1 TB nodes, continue to use this system.
−
If the existing system is the high-performance edition using the 1 TB nodes, continue to use this system.
−
If the existing system is the high-performance edition using the 512 GB nodes, scale up the 512 GB nodes to the 1 TB nodes.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
55
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
6.2 Configuration Calculation of FusionCube for SAP HANA Appliance 6.2.1 Input Information The input information includes:
System scale: either of the following two types of information: − Raw user source data volume and compression ratio
Raw data volume: indicates the amount of data that needs to be imported into the SAP HANA database system for analysis. The customer can provide an estimated data volume based on the data increase requirements. Compression ratio: The raw user source data is compressed and then stored in the memory of computing nodes. The compression ratio is closely related to the data type. An exact data must be provided based on actual tests. The value may range from 3 to 20, and an estimated value can be 5. In some scenarios, the compression ratio can reach 20.
−
Actually required physical memory: indicates the sum of the memory on worker nodes. In practice, this data is obtained based on the raw user source data volume and compression ratio. Sometimes, SAP has already calculated the required physical memory, and the customer can simply provide the value for Huawei. For example, the customer can inform Huawei of providing a 4 TB FusionCube for SAP HANA system. Then, Huawei needs to confirm with the customer whether this data includes the memory of standby nodes. However, in most cases, this data only includes the memory of worker nodes. Size of node: In general SAP consultant will give customer the information about that either 512G node or 1T node will be used. So when customer sends the RFP(REQUEST FOR PROPOSAL) to hardware suppliers he will state which node size will be bought. Number of standby nodes: The default value is 1. The number of standby nodes can range from 0 to N (number of worker nodes).
6.2.2 Calculating number of Computing Nodes Descriptions of the parameters involved in the following formulas are as follows:
Raw-data: specifies the raw user source data volume .
Compression-ratio: specifies the compression ratio.
UP: specifies that the result is rounded up.
Worker: specifies the number of worker SAP HANA nodes.
Standby: specifies the number of standby SAP HANA nodes.
HANA: specifies the number of all computing nodes.
RAM: specifies the calculated total physical memory of worker computing nodes. This meaning applies to all the RAM mentioned in this document. Size: specifies the specifications of each node. For example, the size of a 512 GB node is 0.5 and that of a 1 TB node is 1.
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
56
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
Available formulas are as follows: 1.
Worker = UP [2 x (Raw-data/Compression-ratio)/Size] In this formula: Raw-data/Compression-ratio calculates the amount of user data that needs to be stored and processed in the HANA database. 2 x (Raw-data/Compression-ratio) calculates the actually required physical memory. On each computing node, only a half memory is actually used for storing database data, whereas the other half memory is used for temporary computing purposes. UP: specifies that the result is rounded up. The calculated data may be not an integer. Therefore, to ensure sufficient data storage space, the calculated memory needs to be rounded up. For example, if the calculated memory is 3.09, the value needs to be rounded up to 4. Otherwise, some data will fail to be written into the memory.
2.
HANA = Worker + Standby The number of SAP HANA nodes is the sum of the worker and standby nodes. The standby nodes are determined by the customer, and the default value is 1.
3.
RAM = Worker x Size This formula calculates the total physical memory, that is, the memory for which the customer needs to purchase a license. In most cases, SAP calculates this data and informs the customer. Then the customer purchases hardware systems based on this data.
An example is provided as follows: An enterprise needs to import the financial system data into FusionCube for SAP HANA appliance for realtime analysis. According to statistics collection and estimation of the coming three years' data increase rate, 8.7 TB raw data needs to be processed. The compression ratio obtained by SAP consultants is 3. One standby node is used by default, and 512 GB nodes are used: Worker = UP [2 x (Raw-data/Compression-ratio)/Size] = UP [2 x (8.7/3)/0.5] = 12 HANA = Worker + Standby = 12 + 1 = 13 RAM = Worker x Size = 12 x 0.5 = 6 TB In this way:
SAP informs the customer of purchasing a 6 TB SAP HANA database system. (Customers are charged for SAP HANA database software licenses based on worker nodes, and standby nodes are free for license.)
The customer then purchases a 6 TB FusionCube infrastructure for SAP HANA appliance from Huawei:
If 512 GB nodes are used, the customer needs to purchase a FusionCube for SAP HANA appliance system with 13 nodes (12 worker nodes and 1 standby).
If 1 TB nodes are used, the customer needs to purchase a FusionCube for SAP HANA appliance system with 7 nodes (6 worker nodes and 1 standby).
6.2.3 Calculating NAS Storage space
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
57
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
When NAS is used for Shared volume, the method about calculating the number of disk enclosure and the number of disk is as follows:
In the situation of up to 16 nodes, in despite of how much the RAM (sum of worker HANA node physical memory) is, the number of file engine and the number of disk controller enclosure remain unchanged, only the number of disk enclosure and the number of disks will change with the change of Shared volume space.
Every N8500 disk enclosure can provide up to 24 disks.
In general, 900G 10K SAS disk will be used in NAS device.
Three RAID levels are supported, RAID5/6/10, RAID5 is default.
Thinking of easy expansion in future, every 12 disks will be an expansion unit.
12 disks: RAID5: A hot spare disk in all of 12 disks. A checkout disk in all of 12 disks. The remaining is 10 disks. The actual available disk space is: (900*1000^3/1024^4)*10*95%=7.77T---（95%，a threshold of disk safe space protection. When used disk space is over 95% of all disk space, there will be some unpredicted errors.） RAID6: A hot spare disk. Two checkout disks. The remaining is 9 disks. The actual available disk space is: (900*1000^3/1024^4)*9*95%=6.99T RAID10: A hot spare disk. Every two disks will consist of one RAID1 in the remaining 11 disks. The effective disk number is 5. So, the actual available disk space is: (900*1000^3/1024^4)*5*95%=3.88T
24 disks: RAID5: The actual available disk space is: 7.77*2=15.54T RAID6: The actual available disk space is: 6.99*2=13.98T RAID10: The actual available disk space is: 3.88*2=7.76T
36 disks: RAID5: The actual available disk space is: 7.77*3=23.31T RAID6: The actual available disk space is: 6.99*3=20.97T RAID10: The actual available disk space is: 3.88*3=11.64T
48 disks: RAID5: The actual available disk space is: 7.77*4=31.08T RAID6: The actual available disk space is: 6.99*4=27.96T RAID10: The actual available disk space is: 3.88*4=15.52T
Issue 1.0 (2014-01-26)
Now maximum cluster from Huawei FusionCube SAP HANA appliance is the one with 16 1T nodes, at this time the RAM is 15T, Shared volume is as same as the RAM, so maximum Shared volume is 15T. From the above calculating, 48 disks can meet the disk space requirement in all RAID levels situation, it means that maximum disk enclosure number is 2;
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
58
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
6.2.4 Calculating number of Storage Nodes The calculation of storage nodes is related to the number of HANA nodes, SAP requirements for the capacity of persistent storage media, and storage solution provided by Huawei.
Number of HANA nodes: For the calculation, see section 6.2.2 "Calculating number of Computing Nodes."
SAP requirements for the capacity of persistent storage media: −
−
The minimum capacity of the log volume is 512 GB if HANA node size is greater than or equal to 512 GB. In Huawei solutions, the log volume capacity is 0.5 or 1 times that of the node physical memory. In so doing, always five subracks and two cabinets are deployed regardless of the configurations. The data volume capacity is 3 times that of the node physical memory.
−
The shared volume capacity is 1 times that of the RAM.
Huawei offers the following storage solutions: −
All-SSD solution: SSDs are used as the storage media, regardless of the volume type.
−
All-SAS solution: SAS disks are used as the storage media, regardless of the volume type. Composite storage solution: SSDs are used for log volumes, and SAS disks are used for data volume and shared disks.
− −
NAS storage solution: Shared volume uses NAS storage but Log volume and Data volume use FusionStorage storage. By this time Log volume and Data volume can select any one of the previous three storage types
Detail configutaion of storage nodes in four different storage solutions are as follows: 1.
All-SSD solution The all-SSD solution uses Huawei CH223 blades, each of which offers 2 x 2.4 TB SSDs. The actual available space for each SSD can be calculated as follows: − − −
Unit conversion: Each Huawei SSD provides a capacity of 2,409,073,999,872 bytes, which can be converted in to 2409073999872/(1024 x 1024 x 1024 x 1024) TB. System management consumption: Each 2.4 TB SSD is actually used as six logical disks. Managing each logical disk consumes 20.1 GB. Reliability assurance: To ensure reliability, data is stored with two copies. To ensure performance, the system reserves 5% capacity for other tasks.
According to calculation, the actual available space of each SSD is 0.4103 x 2.4 TB, that is, 0.9848 TB after FusionStorage is used.
The following formula calculates the number of required storage nodes: Storage = UP [RAM x MultipleFactor/(SSDNumber x SSDVolume)]
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
59
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
At least three storage nodes must be configured. If the result calculated by t his formula is less than 3, use 3.
Descriptions of the parameters involved in this formula are as follows:
MultipleFactor: specifies the multiplier factor of the storage capacity, that is, the proportion between the persistent storage capacity and the memory. −
The default value is 5, 1 for Shared volume, 1 for Log volume (According to change of node size this factor will change), and 3 for Data volume.
SSDNumber: specifies the number of SSD card contained in each blade. The default value is 2.
SSDVolume: specifies the actual available storage capacity (0.9848 TB according to the preceding calculation), equivalent to 0.4103 times of the nominal capacity. If FusionStorage is used, each data is stored with 1+1 copies, thereby ensuring data reliability. You can also configure three data copies (1 active and 2 standby) to enhance data reliability. However, this configuration is not recommended except only when the customer has special requirement, because the available capacity is lowered in this mode. RAM: specifies the total physical memory of worker nodes purchased by the user.
UP: specifies that the result is rounded up.
Therefore, still in the preceding example, the following number of storage nodes is required: Storage = UP [RAM x MultipleFactor/(SSDNumber x SSDVolume)] = UP [6 x 5/(2 x 0.9848)] = 16 Altogether 16 CH223 nodes are required.
If 512 GB nodes are used, the customer needs to purchase a FusionCube for SAP HANA appliance with thirteen 512 GB HANA nodes (12 worker and 1 standby). This system consists of 13 HANA computing nodes, 16 all-SSD storage nodes, and one management node into four subracks and two cabinets.
If 1 TB nodes are used, the customer needs to purchase a FusionCube for SAP HANA appliance with seven 1 TB HANA nodes (6 worker and 1 standby). This system consists of seven HANA computing nodes, 16 all-SSD storage nodes, and one management node into three subracks and one cabinet.
2.
All-SAS solution The all-SAS solution uses Huawei CH222 blades, each of which offers 15 x 900 GB/1200 GB 10K SAS disks. Among these disks, two are used as system disks, and others as data disks. A maximum of 13 data disks can be used. However, by default, 12 data disks are configured in most cases. Both 900 GB and 1200 GB SAS disks can be used as data disks, 900 GB is default. The actual available space for a single 900 GB SAS disk is calculated as follows: −
Unit conversion: Each Huawei SAS disk provides a capacity of 900 x 1000 3 bytes, which can be converted in to 900 x 1000 3 /(1024 x 1024 x 1024 x 1024) TB.
−
System management consumption: Managing each disk consumes 20.7 GB. Reliability assurance: To ensure reliability, data is stored with two copies. To ensure performance, the system reserves 5% capacity for other tasks.
−
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
60
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
According to calculation, the actual available space of each 900 GB SAS disk is 0.3792 TB. The actual available space of each 1200 GB SAS disk is 0.5088 TB. The following formula calculates the number of required storage nodes: Storage = UP [RAM x MultipleFactor/(SASNumber x SASVolume)] At least three storage nodes must be configured. If the result calculated by this formula is less than 3, use 3.
Descriptions of the parameters involved in this formula are as follows:
MultipleFactor: specifies the multiplier factor of the storage capacity, which is the same as that in the first solution.
SASNumber: specifies the number of SAS disks on each storage node. The value ranges from 2 to 13, and the default value is 12.
SASVolume: specifies the actual available capacity of the SAS disk.
RAM: specifies the total physical memory of worker nodes purchased by the user. UP: specifies that the result is rounded up.
Still in the preceding example, if 1200 GB disks are used, the following number of storage nodes is required: Storage = UP [RAM x MultipleFactor/(SASNumber x SASVolume)] = UP [6 x 5/(12 x 0.5088)] = 5 Altogether five CH222 nodes are required.
If 512 GB nodes are used, the customer needs to purchase a FusionCube for SAP HANA system with thirteen 512 GB HANA nodes (12 worker and 1 standby). This system consists of 13 HANA computing nodes, five all-SAS storage nodes (each providing twelve 1200 GB SAS disks), and one management node into three subracks and one cabinet.
If 1 TB nodes are used, the customer needs to purchase a FusionCube for SAP HANA system with seven 1 TB HANA nodes (6 worker and 1 standby). This system consists of seven HANA computing nodes, five all-SAS storage nodes, and one management node into two subracks and one cabinet.
Still in the preceding example, if 900 GB disks are used, the following number of storage nodes is required: Storage = UP [RAM x MultipleFactor/(SASNumber x SASVolume)] = UP [6 x 5/(12 x 0.3792)] = 7 Altogether seven CH222 nodes are required.
Issue 1.0 (2014-01-26)
If 512 GB nodes are used, the customer needs to purchase a FusionCube for SAP HANA system with thirteen 512 GB HANA nodes (12 worker and 1 standby). This system consists of 13 HANA computing nodes, seven all-SAS storage nodes (each providing twelve 900 GB SAS disks), and one management node into three subracks and one cabinet.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
61
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
If 1 TB nodes are used, the customer needs to purchase a FusionCube for SAP HANA system with seven 1 TB HANA nodes (6 worker and 1 standby). This system consists of seven HANA computing nodes, seven all-SAS storage nodes, and one management node into two subracks and one cabinet.
3.
Composite storage solution In the composite storage solution, log volumes use the SSDs of the CH223 blades, whereas data and shared volumes use the SAS disks of the CH222 blades. For details about how to calculate the number of CH223 and CH222 blades, see the descriptions of the first and second solutions, respectively. Only the multiplier factors need to be adjusted as follows:
CH223: Storage = UP [RAM x MultipleFactor/(SSDNumber x SSDVolume)] The multiplier factor is 1.
CH222: Storage = UP [RAM x MultipleFactor/(SASNumber x SASVolume)] The multiplier factor is 4.
Therefore, in the preceding example, if the composite solution is used, the following number of storage nodes is required:
Number of CH223 blades: Storage = UP [RAM x MultipleFactor/(SSDNumber x SSDVolume)] = UP [6 x 1/(2 x 0.9848)] = 4
Number of CH222 blades: −
For 1200 GB SAS disks: Storage = UP [RAM x MultipleFactor/(SASNumber x SASVolume)] = UP [6 x 4/(12 x 0.5088)] = 4
−
For 900 GB SAS disks: Storage = UP [RAM x MultipleFactor/(SASNumber x SASVolume)] = UP [6 x 4/(12 x 0.3792)] = 6
Altogether four CH223 and four/six CH222 storage blades are required.
4.
If 512 GB nodes are used, the customer needs to purchase a FusionCube for SAP HANA system with thirteen 512 GB HANA nodes (12 worker and 1 standby). This system consists of 13 HANA computing nodes, four all-SSD CH223 storage nodes, four or six all-SAS CH222 storage nodes, and one management node into three subracks and one cabinet. If 1 TB nodes are used, the customer needs to purchase a FusionCube for SAP HANA system with seven 1 TB HANA nodes (6 worker and 1 standby). This system consists of seven HANA computing nodes, four all-SSD CH223 storage nodes, four or six all-SAS CH222 storage nodes, and one management node into two or three subracks and one cabinet.
NAS storage solution By this time, NAS storage solution is used for Shared volume. Composite storage solution is used for Log volume and Data volume. N8500 NAS provides 1 times of RAM for Shared volume
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
62
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
In the composite storage solution, Log volumes use the SSDs of the CH223 blades, whereas Data volumes use the SAS disks of the CH222 blades. For details about how to calculate the number of CH223 and CH222 blades, see the descriptions of the first and second solutions, respectively. Only the multiplier factors need to be adjusted as follows:
CH223: Storage = UP [RAM x MultipleFactor/(SSDNumber x SSDVolume)] The multiplier factor is 1.
CH222: Storage = UP [RAM x MultipleFactor/(SASNumber x SASVolume)] The multiplier factor is 3.
Therefore, in the preceding example, if NAS storage solution and the composite solution are used, the following number of storage nodes is required:
Number of CH223 blades: Storage = UP [RAM x MultipleFactor/(SSDNumber x SSDVolume)] = UP [6 x 1/(2 x 0.9848)] = 4
Number of CH222 blades: − For 1200 GB SAS disks: Storage = UP [RAM x MultipleFactor/(SASNumber x SASVolume)] = UP [6 x 3/(12 x 0.5088)] = 3 −
For 900 GB SAS disks: Storage = UP [RAM x MultipleFactor/(SASNumber x SASVolume)] = UP [6 x 3/(12 x 0.3792)] = 4
In this example, required RAM is 6T, it means that Shared volume is 6T. If the RAID5 is used and 900G disks are used in NAS, it means total 12 disks and 1 disk enclosure.
Altogether four CH223 and three/four CH222 storage blades are required.
If 512 GB nodes are used, the customer needs to purchase a FusionCube for SAP HANA system with thirteen 512 GB HANA nodes (12 worker and 1 standby). This system consists of 13 HANA computing nodes, four all-SSD CH223 storage nodes, three or four all-SAS CH222 storage nodes, one N8500 NAS with one disk enclosure of 12 disks, and one management node into three subracks and two cabinets.
If 1 TB nodes are used, the customer needs to purchase a FusionCube for SAP HANA system with seven 1 TB HANA nodes (6 worker and 1 standby). This system consists of seven HANA computing nodes, four all-SSD CH223 storage nodes, three or four all-SAS CH222 storage nodes, one N8500 NAS with one disk enclosure of 12 disks, and one management node into two subracks and one cabinet.
6.2.5 Calculation of Networking Connections Still in the preceding example, if the 512 GB all-SSD high-performance solution is used, 13 computing nodes, 16 storage nodes, and one management node are required. System networking The number of networking connection lines depends on the networking design.
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
63
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
By default the computing nodes and storage nodes will be deployed in the same subrack. It means that there are up to 4 computing nodes and up to 4 storage node in one subrack. The reason to do this is that when all 8 computing nodes with 1T size are deployed in the same subrack total power consumption will be over 9KW. Because some customers can’t supply so high power, to assure the availability of power we limit up to 4 computing nodes in one subrack. In chapter 4.4.2, you can find this kind of networking. In this example, because 512G size node is used we can use separate deployment of computing node and storage node.
Figure 6-1 shows the networking in which the computing nodes and storage nodes are deployed in independent subracks. Figure 6-1 Networking connections of FusionCube for SAP HANA
Inter-subrack 10GE connection
Interconnection between two SAP HANA subracks: used for the communication between SAP HANA nodes. To ensure sufficient bandwidth for each SAP HANA node, configure 8+8 optical fibers as a trunk. Cross connections are adopted between two subracks so that the system is still functional when a fault occurs on the inter-subrack network.
Interconnection between the storage subracks and the SAP HANA nodes: Each storage subrack connects to the basic subrack for internal system management. Each storage blade uses two cables to connect to the two 10GE switch modules of the basic HANA subrack.
Therefore, the total number of inter-subrack connections is 20, that is, 16 + 2 x 2. Each optical fiber requires two optical modules. Therefore, altogether 40 optical modules are required.
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
64
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
Inter-subrack IB connection In IB switching, two SAP HANA subracks are used as the aggregation switches. The IB switch modules of each SAP HANA subrack connect to the switch modules of all storage subracks, and each connection uses two cables. Currently, up to 600 MB data is updated on one SAP HANA node per second. Therefore, one SAP HANA subrack requires the following bandwidth: 600 MB x 8 = 4.8 GB = 4.8 x 8 = 38.4 Gbit/s. The IB bandwidth provided by one SAP HANA subrack is 8 x 2 x 56 Gbit/s, which is greater than 38.4 Gbit/s. Therefore, 32 (= 2 x 16) IB cables are required. The IB cables themselves have optical modules, and no external optical module needs to be configured.
NAS connection When N8500 NAS is used for Shared volume there needs 4*10GE and 8 optical modules for storage business lines, 4*GE for management lines. The networking figure is like figure 3-9.
System outlet cables The system provides four pairs of cables for the BI service, data loading, DR replication, SAP HANA management, FusionCube infrastructure management, and SAP HANA client connection. To connect these cables to the switch modules of the E9000 subrack, eight optical modules are required. The peer optical modules are configured by the customer.
6.3 Quick Reference of FusionCube Configurations for SAP HANA Appliance 6.3.1 Typical and Recommended Configurations 6.3.1.1 Typical Configuration According to the configuration principles described in section 6.1 "Configuration Principles of FusionCube for SAP HANA Appliance", the 1 TB node solution of the standard edition based on the Ivybridge platform is used by default.
The standard FusionCube for SAP HANA with four 1 TB nodes (including one standby node) is recommended as the typical configuration, as shown in Table 6-3.
Issue 1.0 (2014-01-26)
System components: four 1 TB HANA nodes, three CH222 storage nodes, one management node, and one E9000 subrack.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
65
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
System characteristics: −
High density: The system with four HANA nodes takes only 12 U space.
−
Large capacity: The available RAM is 3 TB. If the compression ratio is 5, the amount of supported raw data is 7.5 TB.
−
Easy deployment: No internal cable connection is required onsite. Application scenario: This configuration is recommended in the POC test if the customer has no special requirements on the data capacity.
Table 6-2 lists the typical configuration of FusionCube for SAP HANA. Table 6-2 Typical configuration of FusionCube for SAP HANA Westmere EX Platform
Ivybridge EX Platform (E7V2)
(E7V1) 512 GB Node
512 GB Node
1 TB Node
High-Performance Edition
High-Perfo rmance Edition
High-Perfo rmance Edition
Standard Edition
Typical configur ation
Standard Edition Four-node FusionCube for SAP HANA
Table 6-3 lists the specifications of the nodes in FusionCube for SAP HANA in the typical configuration. Table 6-3 Specifications of the nodes in FusionCube for SAP HANA in the typical configuration SNA CH222 (2P+14 x 900 GB SAS+4 GB NVDIMM) SNA CH222 (2P+14 x 900 GB SAS+4 GB NVDIMM) SNA CH222 (2P+14 x 900 GB SAS+4 GB NVDIMM) CH242 V3 (4P+2 x 600 GB SAS+1 TB Memory) CH242 V3 (4P+2 x 600 GB SAS+1 TB Memory) CH242 V3 (4P+2 x 600 GB SAS+1 TB Memory) CH242 V3 (4P+2 x 600 GB SAS+1 TB Memory) MCNA CH222 (2P+2 x 900 GB SAS)
Table 6-4 describes the physical blades in FusionCube for SAP HANA.
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
66
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
Table 6-4 Physical nodes of FusionCube for SAP HANA Blade Type
Description
MCNA
HANA management node: The Management Computing Node Agent (MCNA) is a cloud platform management node providing computing and storage resource pools. FusionManager, FusionStorage Manager, NTP Server and Virtualization Resource Manager (VRM) VMs run on the MCNA node.
SNA
HANA storage node: The Storage Node Agent (SNA) is a host providing computing and storage functions. The FusionStorage software, including the metadata control (MDC) process and object-based storage (OSD) process, run on the SNA node.
CH242 V3
HANA computing node: The CH242 V3 blade is a database host providing computing functions. The SAP HANA databases, file systems, and FusionStorage Agent, including the virtual block store (VBS) process, run on the CH242 V3 blade.
6.3.1.2 Recommended Configuration Recommended configuration The recommended configuration of FusionCube for SAP HANA appliance is the Ivybridge-based standard edition with 1 TB nodes. In the recommended configuration, 1 TB nodes serve as the computing nodes, and CH222 storage nodes use all SAS disks. Each storage node provides twelve 900 GB disks.
Expansion path System configuration and capacity expansion process:
For the first subrack, with start configure at least two computing nodes (including one worker and one standby) and at least three storage nodes. Then you can add the computing nodes to three or four. No storage node needs to be added. If the first subrack is fully configured, it consists of four computing nodes, three storage nodes, and one management node.
For the second subrack, expand the capacity by adding HANA nodes (consisting of one computing node and one storage node) one by one.
For the third subrack, add four storage nodes at a time with the start, and then add HANA computing nodes one by one. The third subrack can contain a maximum of four computing nodes and four storage nodes.
For the fourth subrack, expand the capacity by adding HANA nodes (consisting of one computing node and one storage node) one by one.
The following figures show the sequence of system capacity expansion: 1.
Issue 1.0 (2014-01-26)
From the start to one cabinet with two subracks
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
67
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
->
2.
->
->
From the start on the second cabinet to full configuration
->
->
->
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
68
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
6.3.2 Configuration by Memory – NAS For Shared Volume Because there is a little complex of how to size the system, we give the following tables to get the detail configurations in each user requirements. The customer can determine the configuration based on the RAM (sum of all worker node physical memory) or the number of HANA nodes.
1)
The following configuration is only the recommended minimum configuration which meets the SAP certification and Huawei deployment limit, in actual situations these configurations can be changed according to some special requirements from customer.
2)
When CH222 is used, a FusionStorage storage pool can support up to 96 SAS disks in the condition of two copies. By this time it means up to 8 CH222 blades in a storage pool.
3)
When CH223 is used, a FusionStorage storage pool can support up to 96 logical disks in the condition of two copies. By this time it means up to 16 CH223 blades in a storage pool
4)
When 2-phase 220V power is used, one rack accommodates up to 2 subracks. When 3-phase 380V power is used, one rack accommodates up to 3 subracks.
6.3.2.1 Based on the Ivybridge 1TB nodes: All-SAS disks. Basic information: CPU:
4*Ivybridge 4880
Memory: 32*32G=1T Storage: CH222 blade. Each storage node uses 12 x 900 GB SAS disks. Multiple Factor: Log volume is 0.5 times of Node memory, Node memory, Shared volume is 1 times of RAM
Data volume is 3 times of
The number of Copies: 2 copies for Log volume and Data volume, RADI5 for Shared volume Power: 380V
The following table lists the system configuration based on the Ivybridge(E7V2) 1TB nodes. Table 6-5 System configuration based on the E7V2 1 TB nodes with NAS Shared volume S ystem Scale (RAM ) Number of HANA Nodes ( Including Standby)
1T 2
Issue 1.0 (2014-01-26)
2T 3
3T 4
4T 5
5T 6
6T 7
7T 8
8T 9
9T 10
10T 11
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
11T 12
12T 13
13T 14
14T 15
15T 16
69
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance S ystem Scale (RAM ) Number of CH223 Nodes
1T 3
2T 3
3T 3
4T 4
5T 4
6T 5
7T 6
8T 9(*a)
9T 9
10T 9
11T 9
12T 10
13T 10
14T 11
15T 12
Number of Mgmt Nodes Total Nodes N umber of Subrack s N umber of NAS disks N umber of Cabinet s Power Consumption(W) (*b) Weight (kg) 2 F loor Spa c e （ m ）
1 6 1 12 1
1 7 1 12 1
1 8 1 12 1
1 10 2 12 1
1 11 2 12 1
1 13 2 12 1
1 15 2 12 1
1 19 3 24 2
1 20 3 24 2
1 21 3 24 2
1 22 3 24 2
1 24 4 24 2
1 25 4 24 2
1 27 4 24 2
1 29 4 24 2
4340
5239
6138
8379
9278
10479
11680
14578
15477
16376
17275
19516
20415
21616
22817
439 1.44
454 1.44
469 1.44
605 1.44
620 1.44
650 1.44
680 1.44
1038 2.88
1053 2.88
1068 2.88
1083 2.88
1218 2.88
1233 2.88
1263 2.88
1293 2.88
(*a): When the system needs 9 storage nodes, another storage pool need to be set up, for a new storage pool, minimum 3 storage nodes are needed as a start. (*b): this power consumption is in the data of 80% workload, power consumption data in the all following tables is so.
The blade deployments in every user requirements and their expansion paths are as follows. In figures different color CH222 means that these CH222 belong to different storage pools.
Figure 6-2 Blade deployment and expansion path for 1T all-SAS HANA appliance with NAS
Issue 1.0 (2014-01-26)
->
->
->
->
->
->
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
70
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
->
->
->
->
->
->
6.3.2.2 Based on the Ivybridge 1TB nodes: composite storage Basic information: CPU:
4*Ivybridge 4880
Memory: 32*32G=1T Storage: o
Log volume: CH223 blade. Each CH223 uses 2 x2.4T SSD cards.
o
Data volume: CH222 blade. Each CH222 uses 12 x 900 GB SAS disks.
o
Shared volume: N8500 NAS.
Multiple Factor: Log volume is 0.5 times of Node memory, Node memory, Shared volume is 1 times of RAM
Data volume is 3 times of
The number of Copies: 2 copies for Log volume and Data volume, RADI5 for Shared volume Power: 220V
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
71
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
The following table lists the system configuration based on the Ivybridge(E7V2) in composite storage mode.
1TB nodes
Table 6-6 System configuration based on the E7V2 1 TB nodes in composite storage mode S ystem Scale (RAM ) Number of HANA Nodes ( Including Standby) Number of CH223 Nodes Number of CH222 Nodes Number of Mgmt Nodes Total Nodes N umber of Subrack s N umber of NAS disks N umber of Cabinet s Power Consumption(W) Weight (kg) 2 F loor Spa c e （ m ）
1T 2
2T 3
3T 4
4T 5
5T 6
6T 7
7T 8
8T 9
9T 10
10T 11
11T 12
12T 13
13T 14
14T 15
15T 16
3 3 1 9 2 12 1 6205 582 1.44
3 3 1 10 2 12 1 7104 597 1.44
3 3 1 11 2 12 1 8003 612 1.44
3 3 1 12 2 12 1 8902 627 1.44
3 4 1 14 2 12 1 10103 657 1.44
3 4 1 15 2 12 1 11002 672 1.44
3 5(*b) 1 17 3 12 2 13243 1000 2.88
3 6 1 19 3 24 2 14497 1030 2.88
3 6 1 20 3 24 2 15396 1045 2.88
3 9(*b) 1 23 4 24 2 18241 1211 2.88
3 9 1 24 4 24 2 19140 1226 2.88
4 9 1 26 4 24 2 20314 1253 2.88
4 9 1 27 4 24 2 21213 1268 2.88
4 10 1 29 4 24 2 22414 1298 2.88
4 10 1 30 4 24 2 23313 1313 2.88
(*a): When the number of cluster HANA node is expanded from 7 nodes to 8 nodes, the location of some blades need to be adjusted. (*b): A new storage pool need to be set up.
The blade deployments in every user requirements and their expansion paths are as follows. In figures different color CH222 means that these CH222 belong to different storage pools. When the number of cluster HANA node is expanded from 7 nodes to 8 nodes, the location of one blade which is marked with red need to be adjusted.
Figure 6-3 Blade deployment and expansion path for 1T composite storage HANA appliance with NAS
->
Issue 1.0 (2014-01-26)
->
->
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
->
->
72
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
->
->
->
->
->
->
->
->
6.3.2.3 Based on the Ivybridge 2TB nodes: All-SAS disks. Basic information: CPU:
4*Ivybridge 4890
Memory: 32*64G=2T (*a) Storage: CH222 blade. Each storage node uses 12 x 900 GB SAS disks.
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
73
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
Multiple Factor: Log volume is 0.25 times of Node memory which means that Log volume is 512G, Data volume is 3 times of Node memory, Shared volume is 1 times of RAM The number of Copies: 2 copies for Log volume and Data volume, RADI5 for Shared volume Power: 220V
The following table lists the system configuration based on the Ivybridge(E7V2) 2TB nodes. Table 6-7 System configuration based on the E7V2 2 TB nodes with NAS Shared volume S ystem Scale (RAM ) Number of HANA Nodes ( Including Standby) Number of CH223 Nodes Number of CH222 Nodes
2T 2
4T 3
6T 4
8T 5
10T 6
12T 7
14T 8
16T 9
18T 10
20T 11
22T 12
24T 13
26T 14
28T 15
30T 16
0 3
0 3
0 5
0 6
0 8
0 11(*b)
0 11
0 11
0 13
0 15
0 16
0 19(*c)
0 19
0 20
0 22
Number of Mgmt Nodes Total Nodes N umber of Subrack s N umber of NAS disks N umber of Cabinet s Power Consumption(W) Weight (kg) 2 F loor Spa c e （ m ）
1 6 1 12 1 4628 439 1.44
1 7 1 12 1 5671 454 1.44
1 10 2 12 1 8358 605 1.44
1 12 2 24 1 9756 637 1.44
1 15 2 24 1 11403 682 1.44
1 19 3 24 1 14392 848 1.44
1 20 3 24 1 15435 863 1.44
1 21 3 36 2 16478 1068 2.88
1 24 3 36 2 18125 1113 2.88
1 27 4 36 2 20812 1263 2.88
1 29 4 36 2 22157 1293 2.88
1 33 5 48 2 25146 1459 2.88
1 34 5 48 2 26189 1474 2.88
1 36 5 48 2 27534 1504 2.88
1 39 5 48 2 29181 1549 2.88
(*a): This is not standard SAP HANA node size. Commonly this size is used in the BW (business warehouse) scenario with a condition that CPU is not performance bottleneck. Before using the size, customers should get confirm from SAP. (*b): Another storage pool need to be set up, for a new storage pool, minimum 3 storage nodes are needed as a start. (*c): Another storage pool need to be set up and the location of some blades need to be adjusted so that HANA computing nodes lies only in the first four subracks.
The blade deployments in every user requirements and their expansion paths are as follows. In figures different color CH222 means that these CH222 belong to different storage pools. When the number of cluster HANA node is expanded from 12 nodes to 13 nodes, the location of 10 blades which are marked with red need to be adjusted.
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
74
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
Issue 1.0 (2014-01-26)
->
->
->
->
->
->
->
->
->
->
->
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
75
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
->
->
If customer buys 12 HANA nodes at first, the following deployment will be used. When upgraded to 13 HANA nodes the location of 8 blades need to be adjusted still. In next expansion there is no more location adjust.
->
6.3.3 Configuration by Memory – FusionStorage NFS For Shared Volume Because there is a little complex of how to size the system, we give the following tables to get the detail configurations in each user requirements. The customer can determine the configuration based on the RAM. 5)
Issue 1.0 (2014-01-26)
When CH222 is used, a FusionStorage storage pool can support up to 96 SAS disks in the condition of two copies. By this time it means up to 8 CH222 blades in a storage pool.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
76
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance 6) 7)
When CH223 is used, a FusionStorage storage pool can support up to 96 logical disks in the condition of two copies. By this time it means up to 16 CH223 blades in a storage pool When 2-phase 220V power is used, one rack accommodates up to 2 subracks. When 3-phase 380V power is used, one rack accommodates up to 3 subracks.
6.3.3.1 Based on the Westmere 512GB nodes: All-SSD The following table lists the system configuration based on the Westmere(E7V1) 512 GB nodes. Table 6-8 System configuration based on the E7V1 512 GB nodes S ystem Scale (RAM ) Number of HANA Nodes ( Including Standby) Number of CH223 Nodes Number of Mgmt Nodes Total Nodes N umber of Subrack s N umber of Cabinet s Power Consumption(W) Weight (kg) 2 F loor Spa c e （ m ）
0.5T
1T
1.5T
2T
2.5T
3T
3.5T
4T
4.5T
5T
5.5T
6T
6.5T
7T
7.5T
2
3
4
5
6
7
3 1 6 1 1 3700 370 1.44
3 1 7 1 1 4532 385 1.44
4 1 9 2 1 6679 518 1.44
6 1 12 2 1 8061 558 1.44
7 1 14 2 1 9168 586 1.44
8 1 16 2 1 10275 613 1.44
8
9
10
11
12
13
14
15
16
9 1 20 3 2 12422 936 2.88
11 1 21 3 2 13804 976 2.88
12 1 23 3 2 14911 1004 2.88
13 1 25 4 2 17058 1137 2.88
14 1 27 4 2 18165 1164 2.88
16 1 30 4 2 19547 1204 2.88
19 1 34 5 2 22244 1363 2.88
19 1 35 5 2 23076 1378 2.88
20 1 37 5 2 24183 1405 2.88
6.3.3.2 Based on the Ivybridge 1TB nodes: All-SAS disks. Basic information: CPU:
4*Ivybridge 4880
Memory: 32*32G=1T Storage: Each storage node uses 12 x 900 GB SAS disks. Multiple Factor: Log volume is 0.5 times of Node memory, Node memory, Shared volume is 1 times of RAM
Data volume is 3 times of
Power: 220V
The following table lists the system configuration based on the Ivybridge(E7V2) 1TB nodes. Table 6-9 System configuration based on the E7V2 1 TB nodes S ystem Scale (RAM ) Number of HANA Nodes ( Including Standby) Number of CH223 Nodes Number of Mgmt Nodes Total Nodes N umber of Subrack s N umber of Cabinet s Power Consumption(W) Weight (kg) 2 F loor Spa c e （ m ）
1T 2
2T 3
3T 4
4T 5
5T 6
6T 7
7T 8
8T 9
9T 10
10T 11
11T 12
12T 13
13T 14
14T 15
15T 16
3 1 6 1 1 3915 378 1.44
3 1 7 1 1 4814 393 1.44
3 1 8 1 1 5713 408 1.44
4 1 10 2 1 7954 543 1.44
5 1 12 2 1 9155 573 1.44
6 1 14 2 1 10356 603 1.44
7 1 16 2 1 11557 633 1.44
10(*) 1 20 3 2 14402 989 2.88
10 1 21 3 2 15301 1004 2.88
10 1 22 3 2 16200 1019 2.88
11 1 24 3 2 17401 1049 2.88
12 1 26 4 2 19642 1184 2.88
13 1 28 4 2 20843 1214 2.88
14 1 30 4 2 22044 1244 2.88
15 1 32 4 2 23245 1274 2.88
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
77
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
(*): When the system needs 9 storage nodes, another storage pool need to be set up, for a new storage pool, minimum 3 storage nodes are needed.
6.3.3.3 Based on the Ivybridge 1TB nodes: composite storage Basic information: CPU:
4*Ivybridge 4880
Memory: 32*32G=1T Storage: o
Log volume: CH223 blade. Each CH223 uses 2 x2.4T SSD cards.
o
Data volume and Shared volume: CH222 blade. Each CH222 uses 12 x 900 GB SAS disks.
Multiple Factor: Log volume is 0.5 times of Node memory, Node memory, Shared volume is 1 times of RAM
Data volume is 3 times of
Power: 220V
The following table lists the system configuration based on the Ivybridge(E7V2) in composite storage mode.
1TB nodes
Table 6-10 System configuration based on the E7V2 1 TB nodes in composite storage mode S ystem Scale (RAM ) Number of HANA Nodes ( Including Standby) Number of CH223 Nodes Number of CH222 Nodes Number of Mgmt Nodes Total Nodes N umber of Subrack s N umber of Cabinet s Power Consumption(W) Weight (kg) 2 F loor Spa c e （ m ）
1T 2
2T 3
3T 4
4T 5
5T 6
6T 7
7T 8
8T 9
9T 10
10T 11
11T 12
12T 13
13T 14
14T 15
15T 16
3 3 1 9 2 1 5780 521 1.44
3 3 1 10 2 1 6679 536 1.44
3 3 1 11 2 1 7578 551 1.44
3 4 1 13 2 1 8779 581 1.44
3 5 1 15 2 1 9980 611 1.44
3 6(*a) 1 17 3 1 12221 746 1.44
3 7 1 19 3 2 13422 966 2.88
3 8 1 21 3 2 14623 996 2.88
3 8 1 22 3 2 15522 1011 2.88
3 11(*b) 1 26 4 2 18367 1177 2.88
3 11 1 27 4 2 19266 1192 2.88
4 11 1 29 4 2 20440 1219 2.88
4 12 1 31 4 2 21641 1249 2.88
4 13 1 33 5 2 23882 1385 2.88
4 14(*a) 1 35 5 2 25083 1415 2.88
(*a): When the number of cluster HANA node is expanded from 6 nodes to 7 nodes, or from 15 nodes to 16 nodes, the location of some blades need to be adjusted. (*b): A new storage pool need to be set up.
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
78
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
6.3.4 Configuration by the Number of Subracks 6.3.4.1 NAS for Shared volume The customer can determine the configuration based on the number of subracks. The following table lists the system configuration detail based on the number of subracks when 1TB nodes based on the Ivybridge are used.
All-SAS disk Table 6-11 System configuration based on the number of subracks with 1TB node Item RAM N umber of HANA nodes Number of s torage nod e s N umber of mgmt nodes T ota l n o d e s Number of subracks Number of cabinets
Full Configuration of a Single S u b r a c k 3T 4
Full Configuration o f Two Sub r a c k s 7T 8
Full Configuration o f Three Subracks 11T 12
Full Configuration of a Single Cabinet 7T 8
Full Configuration o f Four Subrack s 15T 16
Full Configuration o f Two Cab i n e t s 15T 16
3
6
9
6
12
12
1
1
1
1
1
1
8 1
15 2
22 3
15 2
29 4
29 4
1
1
2
1
2
2
Composite storage Table 6-12 System configuration based on the number of subracks with 1TB node and composite storage. Item RAM N umber of HANA nodes Number of s torage nod e s N umber of mgmt nodes T ota l n o d e s Number of subracks Number of cabinets
Full Configuration of a Single S u b r a c k N o this situatio n .
Full Configuration o f Two Sub r a c k s 6T 7
Full Configuration o f Three Subracks 9T 10
Full Configuration of a Single Cabinet 6T 7
Full Configuration o f Four Subrack s 15T 16
Full Configuration o f Two Cab i n e t s 15T 16
Minimum two subracks
7
9
7
14
14
1
1
1
1
1
15 2
20 3
15 2
31 4
31 4
1
2
1
2
2
are needed as a start.
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
79
iSite BTS3001C-116 6 Configurations of FusionCube for SAP HANA Appliance
6.3.4.2 Fusionstorage NFS file System for Shared volume
All-SAS disk Table 6-13 System configuration based on the number of subracks with 1TB node Item RAM N umber of HANA nodes Number of s torage nod e s N umber of mgmt nodes T ota l n o d e s Number of subracks Number of cabinets
Full Configuration of a Single S u b r a c k 3T 4
Full Configuration o f Two Sub r a c k s 7T 8
Full Configuration o f Three Subracks 12T 12
Full Configuration of a Single Cabinet 12T 12
Full Configuration o f Four Subrack s 15T 16
Full Configuration o f Two Cab i n e t s 15T 16
3
7
10
10
15
15
1
1
1
1
1
1
8 1
16 2
23 3
23 3
32 4
32 4
1
1
1或2
1
2
2
Composite storage Table 6-14 System configuration based on the number of subracks with 1TB node and composite storage. Item RAM N umber of HANA nodes Number of s torage nod e s N umber of mgmt nodes T ota l n o d e s Number of subracks Number of cabinets
Full Configuration of a Single S u b r a c k N o this situatio n .
Full Configuration o f Two Sub r a c k s 5T 6
Full Configuration o f Three Subracks 9T 10
Full Configuration of a Single Cabinet 9T 10
Full Configuration o f Four Subrack s 13T 14
Full Configuration o f Two Cab i n e t s 15T 16
Minimum two subracks
9
11
11
16
18
1
1
1
1
1
16 2
22 3
22 3
31 4
35 5
1
1
1
2
2
are needed as a start.
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
80
iSite BTS3001C-116 7 Roadmap of FusionCube for SAP HANA Appliance
7
Roadmap of FusionCube for SAP HANA Appliance So far, five solutions of FusionCube for SAP HANA Appliance have been authenticated. For details, see section 5.2 "Specifications of FusionCube for SAP HANA." The future planning is as follows:
Issue 1.0 (2014-01-26)
Larger cluster: Huawei can provide FusionCube for SAP HANA appliance with more than 16 nodes. If the customer poses requirements for more nodes, Huawei can initiate independent product authentication for the customer. Higher cost-effectiveness: Huawei plans to launch the next-generation FusionCube for SAP HANA appliance based on Intel next-generation broadwell platform and to use the computing and storage integrated architecture (requiring no independent storage blades) in 2016 Q4. The next-generation system will offer better performance, higher system density, and higher cost-effectiveness. Keep the synchronization with SAP HANA architecture change: E.g. from SPS09 of SAP HANA database there will be a kind of new node – ES node (Extended Storage). Next-generation FusionCube SAP HANA appliance will support these new functions synchronously.
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
81
iSite BTS3001C-116 8 Acronyms and Abbreviations
8
Acronyms and Abbreviations Acronym and Abbreviation
Full Name
BI
Business Intelligence
BICS
Business Intelligence Consumer Services
BMC
Baseboard Management Controller
BW
Business Warehouse
BWoH
BW on HANA
CRM
Customer Relationship Management
CLI
Command Line Interface
DHT
Distribute Hash Table
DR
Disaster Recovery
ERP
Enterprise Resource Planning
ES
Extended Storage
ETL
Extract, Transform, and Load
FCoE
Fibre Channel over Ethernet
HA
High Availability
IB
InfiniBand
ISV
Independent Software Vendors
JDBC
Java DataBase Connectivity
KPI
Key Performance Indicator
MDX
Multi Dimensional eXpressions
NAS
Network Attached Storage
NFS
Network File System
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
Remarks
82
iSite BTS3001C-116 8 Acronyms and Abbreviations
Acronym and Abbreviation
Full Name
NTP
Network Time Protocol
OCFS2
Oracle Cluster File System
ODBC
Open DataBase Connectivity
ODBO
OLE DB for OLAP
OLAP
Online Analytical Processing
OLTP
Online Transaction Processing
OPS
Operation Per Second
OSCA
Open Service Converged Architecture
A standard used by Huawei E9000 subracks.
PAM
Product Availability Matrix
SAP releases all authenticated products in this lists.
POC
Proof of Concept
RAID
Redundant Arrays of Independent Disks
RPO
Recovery Point Objective
RTO
Recovery Time Objective
SAN
Storage Area Network
SAPS
SAP Application Performance Standard
SI
System Integrators
SoH
Suite on HANA
SPEC
Standard Performance Evaluation Corporation
SQL
Structured Query Language
SSD
Solid-State Drive
TCO
Total Cost of Ownership
TDI
Tailored Data center Integration
XFS
eXtended File System
VM
Virtual Machine
VRM
Virtualization Resource Management
Issue 1.0 (2014-01-26)
Huawei Proprietary and Confidential Copyright © Huawei Technologies Co., Ltd.
Remarks
A kind of Log file system
83
