Transcript
HUAWEI Tecal E9000 Blade Server Pre-sales Specialist Training
Version: V1.0(2014-3-21)
Agenda 1
Click toPositioning add Title Market Overview and
2
Click to add Product Specifications andTitle Highlights
3
Click to add Title Application Scenarios
4
Product Comparison
5
Ordering Guide
6
How to Get Resources 1
X86 Blade Server Market Overview
In 2011, the total server market revenue reached $52.8 billion, among which Unix server revenue reached $17.2 billion (32.6%) and x86 server revenue $35.6 billion (67.4%).
In 2011, the shipment of x86 blade servers took 12.8% (1.195 million PCS) of that of all x86 servers, with the revenue share reaching 21% ($7.53 billion).
In 2014, the total server market revenue will reach $56 billion, among which, the market revenue of x86 servers will be $42 billion (75%), and the shipment of blade servers will take 26% of that of all x86 servers, with the revenue reaching $10.92 billion.
HP, IBM, Dell, and Cisco take 82.2% global market share of blade servers.
Global market share of blade server s in 2010 IBM 25%
HP 47%
Inspur Electronics Groupe 1% Bull 1%
Dawning 1%
Dell Other 10% Vendors 6% Cisco 3% Oracle SGI Hitachi NEC Fujitsu 2% 1% 1% 1% 2%
market share of blade servers in 1.7% 0.9% Global 0.9% 1.8% 2011 HP Total 8.2% 9.6%
43.9%
IBM Total
12.5% 20.5%
others
Blade servers are accepted by more and more customers. The market share of Cisco blade server burgeons and encroaches the market shares of other server vendors. 2
Huawei Blade Server Development History 1st-generation carrier-grade blade server T8000 (out of production) Keep step with the industry and applies to China's Internet industry.
2003
2005
Converged architecture blade server E9000 Consolidate storage, computing, and network resources to meet cloud computing, HPC, and highend enterprise application requirements.
T8000 (out of production) widely substitute the midrange computer platform in the telecom industry and applies to China's Internet industry.
2006
2009
1st-generation enterpriselevel blade server E6000 Provide an efficient, unified computing platform for telecom operation, enterprise applications, and midrange and high-end applications in the Internet industry.
3
2012
2nd-generation enterprise-level blade server E6000H Provide superior computing performance and density, and meet requirements for highdensity and cluster deployments.
E9000 Positioning Market positioning
Provide comprehensive high performance computing support for IT large-scale and high-density deployment industries such as cloud computing, high-end enterprise applications, Internet service platform, and HPC.
4
System Structure
Adopt the fully redundant design for power supply, heat dissipation, management, and switching subsystems based on the 12 U/16 blades architecture; use the latest Intel E5 series CPUs; achieve best balance between performance, density, and energy consumption.
Features
High performance, supporting service development High reliability, ensuring proper running Low power consumption, reducing the OPEX. Easy management, simplifying the O&M
Agenda 1
Click toPositioning add Title Market Overview and
2
Click to add Product Specifications andTitle Highlights
3
Click to add Title Application Scenarios
4
Product Comparison
5
Ordering Guide
6
How to Get Resources
5
E9000 Product Panorama chassis chassis
E9000 chassis
• • • •
Adopts the modular design for compute nodes, storage nodes, switch modules, fan modules, and power supply modules 12U high chassis, providing 8 full width or 16 half width slots Support next 3 generation Intel processors Support evolution of next decade network technology
Computing node
Computing node CH121
CH140
CH220
CH221
CH222
CH240
HW 2S Node HW 2*2P Twin Node FW IO Expansion Node FW IO Expansion Node FW Storage Node High density Ultra high density Large memory Large memory Large memory Large memory high performance Strong expansion Strong expansion Ultra large storage
CH242
FW 4P Node FW 4P Node High performance High performance Ultra large memory Large memory
CH242 V3
FW 4P E7 v2 Node High performance Strong Expansion
Switch module
Switch module CX110
GE switch module
CX116
CX310
CX311
CX317
CX611
CX911/CX912
GE pass-thru module
10GE converged switch module
10GE/FCoE converged switch module
10GE pass-thru module
QDR/FDR switch module
10GE/FC Multiplane switch module
6
E9000 Chassis Front view
Rear view Half width slot
E9000 chassis
Half width half hight slot
Fan module
X2 E4
Switch module 2X
Management module
Power supply module 3X
Full width slot
E1 X3
4E
Chassis specification specification
• • • • •
Installed in a standard 19-inch rack, the E9000 is 12 U high. Its dimensions are: 530.2 mm x 442 mm x 840 mm (H x W x D) An E9000 chassis houses 16 half-width slots or 8 full-width slots Management modules support 1+1 redundancy and management protocols and interfaces such as IPMI 2.0, SOL, SSL, SSH, and web Supports up to 6 Platinum PSMs (AC 3000 W/DC 2500 W) in N+N/N+1 redundancy with up to 95% efficiency under 50% load and supports dynamic management for PSMs in the hibernation state Provides 14 fan modules(3 groups) and every group supports single-fan failures
7
CH121 Half Width Computing Node Front view
CH121 product view
Isometric view
Highlights
• • •
Form factor
Half-width 2-socket computing node
Number of CPUs
1 or 2
Processor model
Intel Xeon E5-2600 or E5-2600 v2 Series CPU
Number of DIMMs
24 DDR3 DIMMs, providing a maximum capacity of 768G
Number of HDDs
2 2.5 inch SSDs or SAS/SATA HDDs
RAID support
Support RAID 0 and 1
PCIe expansion
2 PCIe x16 MEZZ modules 1 standard PCIe x8 full-height half-length card
Operating systems supported
Microsoft Windows Sever 2008/2012 Red Hat Enterprise Linux SUSE Linux Enterprise Server Citrix XenServer VMware ESX ……
Operating temperatures
5℃-40℃
Dimensions
215 mm(8.46 in.)*525mm(20.67 in.)*60.46mm(2.38 in.)
High performance: Half slot provides a dissipation capacity of 700W, supporting 2 E5-2690 processors Cost-effective large-memory: Provides 24 DIMMs at 1.5 times the usual height that apply mainstream granules to double capacity with a total of up to 768 GB DDR3 memory capacity (most cost-effective for large-memory applications) Flexible PCIe expansion: support 1 standard FHHL PCIe card，users can choose proper PCIe acceleration solution for service optimization
8
CH140 Half Width Twin Computing Node Front view
CH140 product view
Top view
Form factor
Half-width 2*2-socket computing node
Number of CPUs
2*（1 or 2 ）
Processor model
Intel Xeon E5-2600 or E5-2600 v2 Series CPU
Number of DIMMs
Each 2S node provides 8 DDR3 DIMMs
Number of HDDs
1 2.5 inch SSD/SAS/SATA
PCIe expansion
Each node provides 2 PCIe x8 MEZZ modules
Operating systems supported
Microsoft Windows Sever 2008 /2012 Red Hat Enterprise Linux SUSE Linux Enterprise Server Citrix XenServer VMware ESX ……
Operating temperatures
5℃-35℃
Dimensions
215 mm(8.46 in.)*525mm(20.67 in.)*60.46mm(2.38 in.)
1*2.5inch HDD SSD
4DIMMs
4DIMMs
Highlights
•
High performance : CH140 support 4 maximum 130W E5-2600 processors(E5-2680, 8 cores at 2.7GHz)
•
Ultra-high computing density: One chassis supports 32*2P notes、64*Romley EP 130W CPU, industry highest density
•
Single node maintenance 9
CH220/CH221 Full Width IO Expansion Node Front view
CH220/CH221 product view
Isometric view
Highlights
• • •
Form factor
Full-width 2-socket computing node
Number of CPUs
1 or 2
Processor model
Intel Xeon E5-2600 or E5-2600 v2 Series CPU
Number of DIMMs
24 DDR3 DIMMs, providing a maximum capacity of 768G
Number of HDDs
2 2.5 inch SSDs or SAS/SATA HDDs
RAID support
Support RAID 0 and 1
PCIe expansion
1 PCIe x16 MEZZ modules 4 PCIe x8 FHHL(CH220)or 2 PCIe x6FHHL(CH221)
Operating systems supported
Microsoft Windows Sever 2008 /2012 Red Hat Enterprise Linux SUSE Linux Enterprise Server Citrix XenServer VMware ESX ……
Operating temperatures
5℃-40℃
Dimensions
423 mm(16.65in.)*525mm(20.67 in.)*60.46mm(2.38 in.)
Strong expansion capability: support standard PCIe card, a maximum of 4 PCIe cards are supported in one full width slot, industry No.1 PCIe expansion capability High performance: full width slot provides a dissipation capacity of 1400W, adoption of FHFL GPU cards can greatly improve HPC application performance Cost-effective large-memory: Provides 24 DIMMs at 1.5 times the usual height, most cost-effective for large-memory applications 10
CH222 Full Width Storage Expansion Node Front view
CH222 product view
Isometric view
Highlights
• • •
Form factor
Full-width 2-socket computing node
Number of CPUs
1 or 2
Processor model
Intel Xeon E5-2600 or E5-2600 v2 Series CPU
Number of DIMMs
24 DDR3 DIMMs, providing a maximum capacity of 768G
Number of HDDs
15 2.5 inch SSDs or SAS/SATA HDDs
RAID support
Support RAID 0,1,10,5,50,6,60, 1GB RAID cache
PCIe expansion
2 PCIe x16 MEZZ modules 1 PCIe x8 FHHL card
Operating systems supported
Microsoft Windows Sever 2008 / 2012 Red Hat Enterprise Linux SUSE Linux Enterprise Server Citrix XenServer VMware ESX ……
Operating temperatures
5℃-40℃
Dimensions
423 mm(16.65in.)*525mm(20.67 in.)*60.46mm(2.38 in.)
Ultra-Large storage: Full width slot provides 15 2.5 inch HDD bays, providing a maximum capacity of 15T, very suitable for applications of big data and distributed computing Flexible PCIe expansion: support 1 standard FHHL PCIe card，users can choose proper PCIe acceleration solution for service optimization Cost-effective large-memory: Provides 24 DIMMs at 1.5 times the usual height, most cost-effective for large-memory applications
11
CH240 Full Width Computing Node Front view
CH240 product view
Isometric view
Highlights
• • •
Form factor
Full-width 4-socket computing node
Number of CPUs
2 or 4
Processor model
Intel Xeon E5-4600/core options: 4,6 or 8
Number of DIMMs
48 DDR3 DIMMs, providing a maximum capacity of 1.5TB
Number of HDDs
8 2.5 inch SSDs or SAS/SATA HDDs
RAID support
Support RAID 0,1,10,5,50,6,60, 512MB/1GB RAID cache
PCIe expansion
2 PCIe x16 mezz modules
Operating systems supported
Microsoft Windows Sever 2008 /2012 Red Hat Enterprise Linux SUSE Linux Enterprise Server Citrix XenServer VMware ESX ……
Operating temperatures
5℃-40℃
Dimensions
423 mm(16.65in.)*525mm(20.67 in.)*60.46mm(2.38 in.)
High performance: full width slot supports 4 E5-4600 processors and 48 DDR3 DIMMs, providing a maximum memory capacity of 1.5TB Large storage: full slot provides 8 2.5 inch HDDs bays, suitable for database applications requiring both computing performance and local storage capacity Cost-effective large-memory: Provides 48 DIMMs at 1.5 times the usual height, most cost-effective for large-memory applications
12
CH242 Full Width Computing Node Front view
CH242 product view
Isometric view
Form factor
Full-width 4-socket computing node
Number of CPUs
2 or 4
Processor model
Intel Xeon E7-4800/core options: 6,8 or 10
Number of DIMMs
32 DDR3 DIMMs, providing a maximum capacity of 1TB
Number of HDDs
8 2.5 inch SSDs or SAS/SATA HDDs
RAID support
Support RAID 0,1,10,5,50,6,60, 1GB RAID cache
PCIe expansion
4 PCIe x8 mezz modules, 1 PCIe x16 FHHL standard card
Operating systems supported
Microsoft Windows Sever 2008 / 2012 Red Hat Enterprise Linux SUSE Linux Enterprise Server Citrix XenServer VMware ESX ……
Operating temperatures
5℃-40℃
Dimensions
423 mm(16.65in.)*525mm(20.67 in.)*60.46mm(2.38 in.)
Highlights
•High performance: full width slot supports 4 E7-4800 processors and 32 DDR3 DIMMs, providing a maximum memory capacity of 1TB •Large storage: full slot provides 8 2.5 inch HDDs bays, suitable for database applications requiring both computing performance and local storage capacity •Flexible PCIe expansion: support 1 standard FHHL PCIe card，users can choose proper PCIe acceleration solution for service optimization 13
CH242 V3 Full Width Computing Node Front view
CH242 V3 product view
Top view
Form factor
Full-width 4-socket computing node
Number of CPUs
2 or 4
Processor model
Full series of Intel Xeon E7-4800/core options: 8 ,10 or 15
Number of DIMMs
32 DDR3 DIMMs, support Intel Max. memory buffer width
Number of HDDs
8(8 HDDs version) or 4(4HDDs version) 2.5 inch SSDs or SAS/SATA HDDs
RAID support
Support RAID 0,1,10,5,50,6,60
PCIe expansion
4 PCIe x16 mezz modules 2 PCIe x16 FHHL PCIe cards, 8HDDs version support 1 PCIe x16 FH ¾ L PCIe card additionaly
Operating systems supported
Microsoft Windows Sever 2008 / 2012 Red Hat Enterprise Linux SUSE Linux Enterprise Server Citrix XenServer VMware ESX
Operating temperatures
5℃-40℃
Dimensions
423 mm(16.65in.)*525mm(20.67 in.)*60.46mm(2.38 in.)
Highlights
•High performance: full width slot supports 4 E7-4800 v2 processors and 32 DDR3 DIMMs, supporting Max. Intel memory buffer width •Large storage: provides 8 2.5 inch HDDs bays, suitable for database applications requiring both computing performance and local storage capacity •Flexible PCIe expansion: support Max. 3 standard PCIe card，supporting GPU and PCIe SSD cards, suitable for large data sets and transactionintensive database like HANA, Oracle DB etc. . 14
E9000 Switch Modules Switch Modules CX116
CX310
CX311
CX317
CX611
GE switch module
GE pass-thru module
10GE converged switch module
10GE/FCoE converged switch module
10GE pass-thru module
QDR/FDR switch module
Uplink : 12*GE and 4*10GE Downlink: 32*GE
Uplink: 32*GE Downlink: 32*GE
Uplink: 16*10GE Downlink: 32*10GE
Uplink: 16*10GE and 8*8GFC Downlink: 32*10GE
Uplink: 32*10GE Downlink: 32*10GE
Uplink: 18*QDR/FDR IB Downlink: 16*QDR/FDR IB
CX911/CX912
Switch Modules
CX110
Highlights
Enables data center TRILL large layer 2 network construction Supports DCB lossless Ethernet to bear FCoE and iSCSI Supports a converged network, virtual paths, and flexible Ethernet and FC configuration 15
10GE/FC Multi plane switch module Uplink: 16*10GE and 8*8GFC Downlink: 32*10GE and 16*8GFC
MEZZ Modules Front view
Isometric view
Mainboard connecter MEZZ Modules
Backplane highspeed connecter Model
Available slot
GE
MZ110
MEZZ1/2
NIC mezz module with 4 x GE ports
10GE
MZ510
MEZZ1/2
CNA mezz module with 2 x 10GE ports
Supports FCoE, and FCoE supports NPIV and NPAR. A physical port can be split up to 4 NPARs.
MZ512
MEZZ1/2
CNA mezz module with 4 x 10GE ports
Supports FCoE, and FCoE supports NPIV and NPAR. A physical port can be split up to 4 NPARs.
ROCE mezz module with 4 x 10GE ports
Support ROCE. FC ports support NPIV.
Models
Type
MZ311
Specification
Multi-IO
MZ910
MEZZ1/2
FC mezz module with 2 x 10GE and 2 x 8G FC ports
InfiniBand
MZ610
MEZZ1/2
HCA mezz module with 2 x 4X InfiniBand QDR ports
MZ611
MEZZ1/2
HCA mezz module with 2 x 4X InfiniBand FDR ports
16
Description
E9000 High Backplane Connections I/O Connections from MEZZ card to switch slot Backplane
I/O Connectors
B
B
C
A
C
A
Half-width Node
B
Backplane
A
8x C
B: MEZZ1 A: LOM
B Mezz1 C Mezz2
半宽刀片 Half-width Node
8x
2X
8x 3X
C: MEZZ2 Full-width Node
1E
8x 4E
I/O Connections Specification Specification
• • • • • • •
The three I/O connectors on the backplane for each half-width compute node are A, B, and C. LAN on motherboard (LOM) uses connector A. The upper mezz module in slot MEZZ1 uses connector B. The lower mezz module in slot MEZZ2 uses connector C The reserved LOM is fixedly connected to slots 2X and 3X. The backplane reserves two SerDes lanes for LOM signals to each slot. MEZZ1 is fixedly connected to slots 2X and 3X The backplane reserves eight SerDes lanes for MEZZ1 signals to each slot. MEZZ2 is fixedly connected to slots 1E and 4E The backplane reserves 8 SerDes lanes for MEZZ2 signals to each slot.
17
Mezz Card and Blade Compatibility Matrix Compatibility Matrix
MZ110
MZ510
MZ311/MZ512
Multi-IO Mezz Module MZ910
CH121
Mezz1 /2
Mezz1 /2
Mezz1 /2
Mezz1 /2
Mezz 2
Mezz 2
CH220
Only Mezz1②
Only Mezz1②
Only Mezz1②
Only Mezz1②
Not Support②
Not Support②
CH221
Only Mezz1③
Only Mezz1③
Only Mezz1③
Only Mezz1③
Mezz 2④ 【Only customized in HPC projects】
Mezz 2④ 【Only customized in HPC projects】
CH222
Mezz1 /2
Mezz1 /2
Mezz1 /2
Mezz1 /2
Mezz 2
Mezz 2
CH240
Mezz1 /2
Mezz1 /2
Mezz1 /2
Mezz1 /2
Mezz 2
Mezz 2
CH242
Mezz1 /2/3/4
Mezz1 /2/3/4
Only Mezz1⑤
Mezz1 /2/3/4
Mezz 2/4
Mezz 2/4
Blade Type
GE Mezz Module
10GE Mezz Module
InfiniBand Mezz Module MZ610①
MZ611①
Matrix Remark
① You are advised to install the InfiniBand mezz modules in Mezz2/Mezz4 (Releated to slots 1E and 4E), for cabling convenience because of the high port density and thick InfiniBand cables. ②By default configuration, the Mezz2 slot of CH220 is already occupied (By the signal connector for standard PCIe card expansion slot) , so only Mezz1 of CH220 can be used to deploy network mezz module. ③By default configuration, the Mezz2 slot of CH221 is already occupied (By the signal connector for standard PCIe card expansion slot) , so only Mezz1 of CH221 can be used to deploy network mezz module by default. ④ But in HPC projects, when used CH221 to support GPGPU/co-processors, customized -configuration by PO of CH221 can be provided, the customized-CH221 can use Mezz2 to deploy IB card. (Note: Customized CH221 need to be tracked by R&D now) ⑤ The new CH242 delivered from 2014.1 can support 1*MZ512 or*MZ311 mezz card(Must be deployed on Mezz1.Old version CH242 don't support MZ512/MZ311).The special feature provided by MZ311 is ROCE, and since that the OS drivers of MZ311 are limit to only few new OS versions,MZ311 is not recommend to deploy when no requirement with ROCE. 18
MEZZ and Switch Module Compatibility -1 Application Scenarios
Customer Requirements on Network Planes
Switch Module
Recommended Slots for Switch Modules
Mezz Module
Remarks
compatibility
1. Conventional Ethernet Networking (Web service deployment, big data clusters, etc)
GE network
2 x CX110
2X/3X
1 x MZ110
10GE network
2 x CX310
2X/3X
1 x MZ510
2. Ethernet and FC storage network (Enterprise DC consolidation, virtualization, database cluster, etc)
Ethernet and FC switching network
2 x CX911 Or CX912
2X/3X
1 x MZ910
Configure GE or 10GE transceiver modules based on customer requirements. If GE electrical ports are required, Huawei provides GE RJ45 electric transceiver modules for option.
Ethernet and FC network with ultra-high reliability, requiring physical redundancy on Ethernet cards and HBA cards)
4 x CX911 Or CX912
2X/3X 1E/4E
2 x MZ910
Configure GE or 10GE transceiver modules based on customer requirements. If GE electrical ports are required, Huawei provides GE RJ45 electric transceiver modules for option.
Network with three physically isolated switching planes: the GE switching plane for management, the 10GE switching plane for services, and the FC switching plane for storage
2 x CX911 Or CX912
2X/3X
1 x MZ910
The 10GE and FC switching planes are provided.
2 x CX110
1E/4E
1 x MZ110
The GE switching planes are provided.
19
MEZZ and Switch Module Compatibility -2 Application Scenarios
Customer Requirements on Network Planes
compatibility
Switch Module
Recommended Slots for Switch Modules
Mezz Module
3. Converged Ethernet Converged Ethernet networking (VDI, enterprise DC with IP SAN consolidation, Converged Ethernet networking virtualization, etc) with FC SAN
2 x CX310
2X/3X
1 x MZ510
Multiple-plane communications can be implemented by VLAN.
2 x CX311
2X/3X
1 x MZ510
Multiple-plane communications can be implemented by VLAN.
4. InfiniBand network (HPC)
1 x CX110
2X/3X
1 x MZ110
If Ethernet redundancy is required, configure one more CX110.
1 x CX610
1E/4E
HPC network: GE + InfiniBand network, non-redundancy plane
20
Remarks
1x Choose the MZ610 (QDR) or MZ611 (FDR) based MZ610/MZ611 on customer requirements.
E9000 MM910 Management Module Main parts and ports
FPGA
DSP
Management Module
CPU
Local KVM port
Management GE port Management serial port
Description description
•
The MM910 manages the E9000, complies with IPMI 2.0, and provides management functions such as remote startup, shutdown, reset, logging, hardware monitoring, SOL, KVM over IP, virtual media, fan monitoring, and PSM monitoring.
•
The MM910 supports 1+1 redundancy, and provides a local KVM port for server management.
21
E9000 Converged Infrastructure Blade Server 1 2
3 •
Computing 7 kinds of computing nodes，fully support E5-2600/E5-4600/E74800 processors
•
Storage Distributed storage solution provides cost-effective solution for applications of big data
•
Network 8 kinds of switch modules support GE/10GE/FCoE/FC/IB QDR/FDR
•
Management Esight unified management improves management efficiency over 70%
22
No.1 32*2P notes and 64*Romley EP 130W CPU per chassis Computing Maximum floating point performance 16.5TFlops / chassis Density No.1 Storage Density
CH222（Full width）support 15*2.5 inch HDDs Single-frame support 120*2.5 inch HDDs
E9000 switch module base on advanced Huawei DataCenter Switch Technology 15.6T backplane switch capacity，support maximum 128*10GE Performance ports and evolution to 40GE and 100GE Support Ethernet，IB，FC ports, flexible networking
No.1 Switch
4
No.1 Reliability
Support running at 40°C The failure rate is 15% lower than competitors
5
No.1 Energy Efficiency
Rank first in China Mobile and China Unicom competition testing Chinese green energy certification
6
Plug and Play
Convenient Automatic configuration and transmission data, including O&M BIOS, RAID, network card
No.1 Computing Density E900 Highest Computing Density
One chassis supports 32*2P notes、64*Romley EP 130W CPU Maximum floating point performance 16.5TFlops Every node supports 8 memory channels, leading memory bandwidth
HUAWEI CH140
Quantity of CPUs / E5-2600 Single node maintenance Quantity of memory channel
DELL Power420
Half width slot support 2*2P nodes, E5-2600 CPU，every node support 8 memory channels, support single node maintenance
Full hight 4 nodes support E5-2400 CPU, every node only support 6 memory channels. Not support single node maintenance
23
HUAWEI E9000
DELL M1000e
IBM FlexSystem
64/12U
32/10U
28/10U
Yes
Yes
No
8
6
6
IBM FlexSystem X222 full width slot 4 nodes support E5-2400 CPU， every node only support 6 memory channels
No.1 Storage Density E9000 Highest Storage Density
E9000 CH222（Full width）support 15*2.5 inch HDDs
Quantity of HDDs per slot
Single-frame support 120*2.5 inch HDDs，the biggest memory
Quantity of HDDs per chassis
HUAWEI E9000
DELL M1000e
IBM FlexSystem
15
7
14
120
56
98
capacity in one chassis Note: E9000 chassis supports 8 CH222 computing nodes
HUAWEI CH222
DELL internal SAN PS-M4110x
E9000 CH222：Support 15*2.5’ HDDs，SAS, SATA, NL-SAS, 120 HDDs in one chassis
Dell SAN PS-M4110x occupies double slots， support 14 HDDs， 56 HDDs in one chassis
24
IBM FlexSystem storage node IBM Flexsystem storage note, support 14*2.5’ HDDs， 98 HDDs in one chassis
No.1 Switch Performance E9000 switch performance overall leading
E9000 switch module base on advanced Huawei DataCenter Switch Technology 15.6T backplane switch capacity，support maximum 128*10GE ports and evolution to 40GE and 100GE Support Ethernet，IB，FC ports, flexible networking
Backplane bandwidth Quantity of 10GE
E9000 Switch Module Switch slot uplink 32 ports，single-frame maximum 128*10GE ports， support more than 3 plane network
EN4093 10Gb switch board： 14*10G+2*40G
25
IBM FlexSystem
HP C7000
15.6T
7.292T
7.168T
6.144T
128
48
88
64
IBM: single-frame max 88*10GE ports，only 2 network plane HP/DELL：switch module uplink 16*10GE ports
CX317 10GEpass-thru: 32ports
CX911 multi-plane switch board
DELL M1000e
Other Manufacture
EN4091 10Gb Board： 14 ports CX116 GE pass-thru: 32ports
HUAWEI E9000
HP 6120XG 10G switch board: 8 ports
10G switch board ：8 ports
IB switch board ：8ports
No.1 Reliability Ensure Reliability by Advanced Design and Strict Testing Support stable operation at 40°C high temperature, ensuring business continuity The failure rate is 15% lower than competitors Independent air-duct, improving system heat dissipation
Passive backplanes avoid single points of failure •
•
Front and rear independent air-duct
•
Duct design optimization
The passive backplanes contain no active components such as resistors, capacitors, and ICs, except PCBs and connectors
Failure Rate(According to the statistical result from customer T)
The most stringent standards of reliability testing for components • • •
High and low temperature test, HALT 3000 times Power cycle test Continuous 30 days stress test
Others
0.87%
1.12%
Quality control throughout the whole process • • • • •
26
HUAWEI
Supplier Certification Component selection and control IPD process from module to system 150 reliability patents EMC testing, safety testing
No.1 Energy Efficiency Leading Energy Efficiency, Energy Saving
Rank No.1 in SEPC power testing
Rank No.1 in in China Mobile and China Unicom competition testing 80PLUS Platinum Power efficiency 94.6%
• •
80PLUS Platinum power, 94.6% conversion efficiency onboard DC/DC 94% conversion efficiency
Chinese Green Energy Certification • •
Non-toxic design, Energy saving Green energy, to meet customer requirements of environmental protection
27
Chinese green energy certification
DEMT--Dynamic Energy Management Technology
Power Capping Technology • •
Improve power utilization of the equipment room and cabinet Improve equipment room density, reduced rental costs;
• •
Improve energy efficiency over 13% by DEMT Range of energy saving cover idle～100％ load
Convenient O&M Initial deployment
•
Deployment, changing or migration of blades or switch modules
Offline Configuration
Equipment changing
Configuration Remains
Equipment migration
Configuration Migration
Mass deployment
Plug and play
Configuration migration with equipment location
• Plug and play • Parameters automatically configured and migrated (including configurations of BIOS，network card，RAID etc.)
28
•
Time-saving, changing equipment would only cost 2 minutes instead of 20 minutes
Technical Innovation: Accelerating Service Development Hardware acceleration solutions greatly accelerate services.
Use the PCIe SSD card to improve the IOPS by 100 times.
Use the coprocessor to provide an acceleration ratio of 300%.
Use the compression card to improve the Hadoop performance by 40%.
Use the iNIC to improve the Vswitch performance by 200%.
Huawei 4T financial data query test
Top 17 Financial Ledger SQL Query
SSD card
Exadata
Huawei
SQL17
Timeout
320s
SQL13
905s
319s
SQL12
737s
124s
SQL7
85s
15s
iNIC
CH221 I/O expansion compute node E9000 chassis
Intel Phi coprocessor
29
Differential Advantage: Huawei Large-Capacity DIMMs at 1.5 Times the Usual Height DIMMs at 1.5 times the usual height are used in virtualization solutions, reducing memory investment by 20% to 30%.
DIMMs at 1.5 times the usual height
Huawei
1.5 times the usual height 42 mm to 46 mm (1.65 in. to 1.81 in.)
Universal DIMMs in the industry
Standard height 30 mm (1.18 in.)
• IBM
VLP 19 mm (0.75 in.)
•
30
The RH2288H V2 supports DDR3 DIMMs at 1.5 times the usual height, reducing costs by 20% to 30% compared with DIMMs at the usual height (in the case of 32 GB memory). All E9000 compute nodes, except for the CH140 and CH242, support DIMMs at 1.5 times the usual height.
Differential Advantage: DEMT Energy saving statistics of the RH2288 V2 using the dynamic energy management technology (DEMT) Energy saving statistics
Remarks: The data is obtained for the RH2288 V2 server in the integer computing benchmark test. The effect varies depending on server configurations. DEMT
Description
The DEMT is a dynamic energy-saving technology developed by Huawei. The technology automatically monitors server resource usage based on service loads, dynamically adjusts server running status based on resource utilization, provides minimum power based on service requirements, and minimizes server power consumption. • Slashes system power consumption in various load scenarios without affecting service performance. • Automatically triggers and monitors energy saving in closed-loop mode without manual intervention. • Lowers power consumption for PSUs, processors, and fans based on service requirements.
31
Differential Advantage: Power Capping Technology Background Background and benefits
• • •
Benefits
The power supply and heat dissipation costs for data centers dramatically increase and even exceed device costs. The power supply and heat dissipation capabilities of data centers do not meet requirements for service expansion. The power supply and heat dissipation equipment of data centers is underused.
• • • •
Assigns power supply and heat dissipation resources based on actual device power, improving energy usage. Increases reliability without affecting service performance. Eliminates the risks of insufficient heat dissipation and power supply if air conditioners are faulty or service loads increase. Supports more devices in a rack.
Value
Test data A rack can house more devices after the power capping technology is used.
32
Differential Advantage: Huawei Proprietary Heat Dissipation Design Key technologies
Key technologies • • • •
Vector airflow management: ensures heat dissipation for high-power processors (95 W/130 W/135 W) with efficient utilization of limited airflow. Double-faced cellular board: increases the amount of system airflow input by achieving the open porosity of 43%, an increase of 115% compared with HP servers. Multiple-input multiple-output (MIMO) dynamic energy control: reduces energy consumption by 4.22%, raising the ranking from No.6 to No.1. Aluminum board embedded with copper heat sink: reduces the heat sink weight by 45%, cutting the cost by 40%.
Huawei proprietary heat dissipation design
Aluminum board embedded with copper heat sink:
Description
Benefits
System vector airflow management
E9000 compute node
MIMO E9000 chassis Double-faced cellular board
• Ranks No. 1 in the SPECpower test. • Supports smooth product evolution and expansion and supports architecture evolution in 10 years and processor evolution based on three generations. • Supports operating temperature 40°C (104°F), which is higher than the normal operating temperature 35°C (95°F).
The excellent heat dissipation design of the E9000 supports long-term stable operating at 40°C (104°F) and technology evolution in 10 years and processor evolution based on three generations.
33
Differential Advantage: MIMO MIMO principles and test data
Diagrams
Temperature (chips, Ta) CPU states Fan speed
System power
Output
Input
Fan speed CPU states
Fan control
Target temperature
Noise limitation
CPU high utilization
Relationship between the processor leakage current and temperature
Temperature limitation
CPU low utilization
Fan speed Relationship between the optimal energy saving point and fan speed
MIMO heat dissipation and energy saving technology Description
•
• • •
When the traditional fan speed adjustment technology is used, the fan speed is minimized when the device is operating at the normal temperature, ensuring lowest noise while implementing heat dissipation. As the x86 manufacturing process develops, the leakage current of processors increases with the temperature. For example, when the temperature of a processor increases by 1°C (33.8°F), the power consumption of the processor increases by about 0.7 W. The fan power consumption is proportional to the fan speed by cube. The MIMO technology dynamically controls the fan speed in real time, balances the fan power consumption and hardware power consumption, and optimizes the system energy consumption based on parameter value changes. Input parameters for energy saving control include the temperature (chip temperature and ambient temperature), processor running status, fan speed, and thresholds (noise/temperature). Output parameters include the component target temperature, fan speed, and processor status. The MIMO technology saves energy by 5%, which is higher than the traditional fan speed adjustment technology.
34
Differential Advantage: Double-faced Cellular Board Diagrams
Diagrams
Porosity rate 58.9%
Porosity rate 75%
Double-faced cellular board Description
• • •
Compared with other structure, closed equilateral hexagonal honeycomb structure achieves the maximum force with the minimal materials and maximizes the porosity area to ensure the maximum cooling area with the proper strength. Large-area bionic cellular heat dissipation holes ensure proper panel strength, maximize the cooling area, and ensure high air intake efficiency. The most air intake volumes in unit time bring most cool air, which ensures excellent cooling performance for servers. Huawei servers use double-faced cellular board design, which adopts cellular heat dissipation holes on both sides of disk trays and provides lower air resistance than competitors' products.
35
Differential Advantage: Shock Absorption Design Diagrams
Diagrams Shock absorption design Description
•
Shock risks in expected application environments are assessed during system design. Proper shock absorption is designed to prevent shock failures.
•
Packaging buffer design ensures safe device transportation.
•
Reasonable fan speed adjustment and fan component shock isolation are designed to avoid HDD-sensitive frequencies and to decrease risks of HDD performance deterioration caused by rotating components.
36
Differential Advantage: High-Standard EMC Design Description
Advantages of Huawei EMC design •
20-year experience in electromagnetic immunity designs for Huawei network products in complex application environments
•
General-purpose, high-reliability surge protection design
•
Strict RE control based on global standards Huawei EMC competitiveness comparison EMC Competitiveness Comparison Huawei Conducted susceptibility
10 V
Pictures and data
Signal port EFT 1 kV Power supply port 2 kV EFT Radiated 10 V/M susceptibility ESD susceptibility Anti-surge capability Radiation emission (RE)
Competitor 3V 0.5 kV 1 kV 3 V/M
Contact discharge 6 kV Air discharge 8 kV
Contact discharge 4 kV Air discharge 8 kV
ITU 2.5/2.5 kV
IEC 2 kV/2 kV
CE/FCC Class A
CE/FCC Class A
Huawei servers provide higher anti-interference and surge-protection capabilities than competitors' products, supporting high adaptability.
37
Differential Advantage: Huawei-developed Technologies and Optimization Capability Huawei-developed technologies and optimization capability
Description • •
•
• •
Architecture and solution design Component selection
•
Schematic diagram
•
PCB
•
Heat dissipation simulation Signal integrity simulation
• •
BIOS and BMC CPLD and FPGA logic
•
iMana (management software) - substitute for USM
Most competitors in China do not have comprehensive R&D systems. Most of them provide ODM servers purchased from Taiwan and United States. Competitors such as HP, IBM, Dell, and Cisco design architecture and management software and partner with ODM vendors such as Foxconn, Inventec, Quanta for design, test, and production.
38
Smooth Evolution: Supporting Network Evolution in 10 Years Technology development trend
Diagrams Network development trend and E9000 architecture planning With the emergence and popularization of 100GE, InfiniBand EDR, and 32G FC technologies in the next
Description
few years, support for signals at a rate higher than 25 Gbit/s is essential for smooth network evolution. The HUAWEI E9000 high-speed chassis is designed to support 100GE and InfiniBand EDR evolution. To help customers maximize return on investment (ROI) in the future 10 years, proper architecture planning is implemented to ensure that the E9000 supports high-speed signals and high-speed backplane signal links. This meets requirements for network evolution in the future 10 years and provides competitive advantages for customers in the big data analysis market by delivering high performance, high bandwidth, and low latency.
39
Smooth Evolution: Supporting Processor Evolution in 10 Years Independent air channel design and Intel processor planning Processor (Tock)
Processor
Processor Type
Processor with the
Release Date
MAX TDP
Highest Power Romley
Ivy-Bridge
Pictures
IVB-EP 2S
E5-26XX V2
2013 Q3
135 W (plan)
IVB-EP 4S
E5-46XX V2
2013 Q4
130 W (plan)
IVB-EX
E7-48XX V2
2013 Q3
155 W (plan)
E7-88XX V2 Grantly
• •
Separated air channels of the server blades, • switching subsystem, and power supply subsystem • Intelligent fan speed adjustment and high efficiency heat dissipation
Air vents in the backplane ensuring the shortest air channels for nodes 1.5U space allowing for large air intake space
Haswell
HSW-EP 2S
E5-26XX V3
2014
145 W (plan)
HSW-EP 4S
E5-46XX V3
2014
135 W (plan)
HSW-EX 2S
E7-48XX V3
2014
165 W (plan)
E7-88XX V3 Next Generation
> 165 W?
E9000 heat dissipation specifications and design for processor evolution based on three generations
Description
• • • • •
Wide operating temperature range: Compute nodes' longterm operating temperature ranges from 0°C to 40°C (32°F to 104°F). 700 W heat dissipation for a half-width compute node slot 1400 W heat dissipation for a full-width compute node slot 280 W heat dissipation for a switch module slot Compute nodes supporting the Intel E5/E7 processor family work efficiently when DIMMs are fully configured.
40
More cores need to be integrated into a processor due to virtualization applications of data centers and competition requirements, especially competition between the RISCtolA and the midrange computer. As a result, processor power consumption is still on the rise. From the released Intel road map, the maximum power consumption of processors for servers increases to 155 W and 165 W. The HUAWEI Tecal E9000 meets the predicted requirements by providing excellent heat dissipation design for compute nodes, which facilitates applications of processors of 165 W or higher.
Why the Huawei E9000 Uses a 12U Chassis? Advantages of a 12U chassis
Description
12U is the optimum height designed based on computing density, computing and storage capabilities, and long-term evolution. 12U helps customers maximize ROI and minimize TCO by providing a wide range of hardware expansion resources. 12U has the following advantages: • With 12U height, the E9000 chassis has sufficient heat dissipation capabilities, can operate properly for a long term at 40°C (104°F) to adapt to more severe environments, and follows the processor development trend and overall component power consumption development trend of the next six or eight years to help customers maximize ROI. • 12U provides 2.4-inch height per slot so that each slot can house 2.5-inch hard disks in three drawers and 3.5-inch hard disks in two drawers. This greatly expands system storage capacity to better support integration with distributed storage and big data applications. • 12U space can house 16 half-width slots and 8 full-width slots. Each node supports DIMMs at 1.5 times the usual height. Conventional blade server vendors reduce the chassis height to increase computing density per unit space. Such design leads to the use of standard-height DIMMs and VLP DIMMs, which are less cost-effective than DIMMs at 1.5 times the usual height. Note： Note: If a competitor promotes 10U height with four chassis in a rack during bidding, attempt to purchase the customer as follows: The three chassis in an E9000 rack provide higher computing density (CH140) than the four chassis in the competitor' rack, and reserve 4U space for placing storage devices, switches, rack servers, and other devices.
41
Agenda 1
Click toPositioning add Title Market Overview and
2
Click to add Product Specifications andTitle Highlights
3
Click to add Title Application Scenarios
4
Product Comparison
5
Ordering Guide
6
How to Get Resources 42
Application Scenarios Enterprise Datacenter
Physical machine, VM, and infrastructure with converged computing, storage, and networking resources
Virtual Desktop
VDI in a BOX
Appliance
HPC
Key Application
Application performance acceleration， High-speed Fabric interconnection， Business intelligence/DW, memory database, big data
Four-socket compute nodes supported in half-width slots, InfiniBand switching
RISC to IA
The modular converged architecture meets various workload requirements Computing
2S, 4S computing nodes, storage expansion, IO expansion computing nodes
Storage
IPSAN, FC-SAN, FCoE, distributed storage
Network
GE, 10GE, 8GFC, FCoE, Infiniband switch, TRILL large layer 2 data center switches
IO Acceleration components
PCIe SSD, GPU
43
Enterprise Data Center Application Physical and Virtual Servers Enterprise Datacenter Application
Typical Configuration • E9000 chassis • CH121 2-socket compute node (for web, middleware, and application VM
LAN
deployment)
GE
• CH220 resource expansion 2-socket compute node • CH222 storage expansion 2-socket compute node • CH240 4-socket compute node (database physical machine deployment)
10GE 8*8G FC
10GE LAN
• Large-capacity SSD cards with application acceleration for email, ERP, CRM,
8G FC SAN
databases, and BI • CX110 and CX311 switch modules
iSCSi FCOE
IPSAN FCSAN
• GE, 10GE, FCoE, and FC switching, providing a converged network to improve management convenience, flexibility, and scalability. FC SAN
44
VDI in a Box Computing and Storage Convergence Without SAN VDI in a box
Typical Configuration • • • •
E9000 chassis CH222 storage expansion 2-socket compute node CX110 switch module PCIe SSD card
Features
GE/10GE LAN
• • • • •
45
Supports 128 processor cores, 6 TB memory capacity, and up to 128 TB raw storage capacity. Supports 500 (heavy) to 1000 (light) virtual desktop infrastructure (VDI) users. Supports deployment of an OS and hot applications on PCIe SSD cards, which reduces the VDI startup peak, increases IOPS by 800%, and reduces costs by 20%. Supports the disk drawer and hot-swappable disks to reduce the time required to rebuild data during disk replacement. Supports the virtual block storage (VBS) distributed harsh table (DHT) ring to protect shared disk storage in cross-server mode.
MPP Architecture OLAP for BI, DW, and In-memory Database MPP Architecture
Typical Configuration • E9000 chassis • CH222 storage expansion 2-socket compute node • CH240 4-socket compute node (1.5 TB memory capacity) • PCIe SSD card • CX311 switch module
10GE LAN
Features • Provides 2-socket 24-DIMM computing capability per node. Supports 15 disks per node. • A 4-socket 48-DIMM analysis node provides 1.5 TB memory capacity to support the memory database and real-time analysis. • Each node provides service bandwidth of 20GE. The switch latency is in ns, which removes the IO bottleneck during the ETL process. • The large-capacity PCIe SSD card is used for the index and hot data cache, which greatly improves the IOPS of the parallel database.
56G IB
46
Unstructured Database Application Hadoop, Hbase Application Unstructured Database Application
Typical Configuration • E9000 chassis • CH121 2-socket compute node • CH222 2-socket storage expansion compute node (for web, middleware, application VM deployment) • Hadoop hardware compression card • CX110 switch module
GE LAN
Features •
10GE
•
•
47
A NameNode 2-socket compute node can provide a cost-effective large memory with up to 768 GB memory capacity. A DataNode 2-socket compute node supports 15 x 2.5" disks, and provides 40GE service bandwidth (4GE per node). The compression card improves Hadoop performance by 20%, and reduces network traffic by more than 60%, CPU usage by over 80%, used disk space by more than 50%, and disk I/O data read and write operations by over 50%.
Key Applications RISC to IA, Unix Migration Key Applications 32 x 10GE uplink ports LAN
Typical Configuration
12 x GE uplink ports
16 x 10GE uplink ports
• E9000 chassis • CH121 2-socket compute node (for web, middleware, and application VM deployment) • CH240 4-socket compute node (database physical machine deployment) • CH280 8-socket compute node (2013) • PCIe SSD card • CX110 and CX311 switch modules
Features
8 x GE downlink ports
8G FC SAN FC
8 x GE downlink ports
48
• Supports the VM resource pool for web, application, and middleware layers and virtual NICs to increase network throughput, which ensures VM security. • The OLTP database is deployed on physical machines, which supports stateless computing and migration, and use FC SAN for shared storage. • Supports a large-capacity cache by using PCIe SSD cards to improve I/O performance for databases. • Supports highly reliable redundancy for PSMs (N+1/N+N) and fan modules.
High Performance Computing HPC Application HPC
Feature
• CH140 4S high density computing node
CH221 IO expansion 2S computing node
49
CH140 Twin Computing Node: High-density computing
•
Supports 2*2-socket compute nodes in half-width slots
•
Supports 64 Intel Xeon E5-2600 processors in a chassis.
•
Supports 192 E5-2600 processors in a rack(3 chassises).
•
Supports 12 TB memory capacity
•
Supports InfiniBand QDR/FDR
•
Supports InfiniBand EDR evolution
•
Supports GPU and SSD resource expansion
Case: Dispatching and directing management system for CREC Customer requirements
As the dispatching and directing management system is closely related to production security, the China Railway Engineering Corporation (CREC) requires a system with high reliability and security. The dispatching and directing management system frequently transmits data, which imposes a high requirement on the data transmission latency. The system must be easy to maintain.
Highlights
50
Reliability: The E9000 uses a passive backplane and fully redundant key components, which greatly improves system reliability. Real-time capability: The E9000 backplane provides a switching capability of up to 15.6 Tbit/s, ensuring high-speed data switching for internal service modules. Maintainability: The E9000 provides unified eSight management interfaces for configuring, monitoring, and managing the chassis, PSUs, fan modules, switch modules, and compute nodes.
Case: HPC for UK XX University Customer requirements • • •
Storage: The distributed storage solution is used to allow all compute nodes to share the storage space. Computing: The system supports the next-generation Intel CPUs with the data processing capability of 8 TFLOPS at peak values. Network: The multiple-plane networking is used to isolate the computing, storage, and management networks.
Solution • •
• 51
The Lustre distributed storage solution is used, and compute nodes and storage nodes are selected for the Lustre networking. The existing platform can be smoothly upgraded to Ivy Bridge, which meets requirements for the architecture and computing performance. The GE, 10GE, and IB networking solutions are provided, and physical planes are isolated, which ensures stable service operating.
Agenda 1
Click toPositioning add Title Market Overview and
2
Click to add Product Specifications andTitle Highlights
3
Click to add Title Application Scenarios
4
Product Comparison
5
Ordering Guide
6
How to Get Resources
52
E9000 Chassis Advantages
Advantages
E9000 chassis advantages Switching
Industry-leading backplane switching capability of 15.6 Tbit/s
Storage
8 CH222 storage nodes, configured with 120 x 2.5" internal hard disks in a chassis for providing the largest internal storage capacity in a chassis.
Computing
64 Romely EP 130 W CPUs in a chassis, with the highest density in the industry
Heat dissipation
PSU
53
Support for stable operating at 40°C Note: A.Only IBM servers support the 40°C specifications among competitors. B.The CH140 twin compute nodes support the temperature range from 5°C to 35°C, and therefore the CH140 twin compute nodes support CPUs of up to 130 W. 80 Plus Platinum PSUs with power efficiency of 94.6%, which is 5% higher than the industry level
E9000 Compute Node Advantages Compute Node Half-width CH121
Advantages
Half-width CH140 twin compute nodes
Form Factor
CPU
Memory
Storage
DIMMs at 1.5 times the usual height
N/A
One PCIe x8 FH HL SSD on each compute node
Independent node maintenance
64 x 130 W E5-2600 CPUs in a chassis, with the highest density in the industry
N/A
N/A
DIMMs at 1.5 times the usual height
N/A
Full-width CH221 I/O expansion node
N/A
N/A
DIMMs at 1.5 times the usual height
N/A
Full-width CH222 storage node
N/A
N/A
DIMMs at 1.5 times the usual height
15 x 2.5" hard disk slots in a full-width slot of 1.5U
DIMMs at 1.5 times the usual height
8 x 2.5" hard disk slots (The number of hard disks supported on a 4-socket compute node is twice that of a conventional 4-socket compute node.)
Full-width CH220 I/O expansion node
Full-width, 4-socket E54600 CH240 compute node
Full-width, 4-socket E74800 CH242 compute node
N/A
N/A
N/A
N/A
54
N/A
N/A
PCIe Expansion
N/A
8 x 2.5" hard disk slots (The number of hard disks supported on a 4-socket compute node is twice that of a conventional 4-socket compute node.)
N/A
4 PCIe x8 FH HL slots for installing 4 GPU or PCIe SSDs per compute node N/A One PCIe x8 FH HL SSD on each compute node
N/A
One PCIe x8/x16 FH HL slot
Competitiveness Comparison Between the E9000 and Products from Tier 1 Competitors Key Competitiveness
Key Feature
HP C7000 G8
IBM FlexSystem
Cisco UCS G3
Dell M1000e
HUAWEI E9000
1200 W/100 W
1200 W/100 W
1200 W/100 W
1400 W/280 W
1400 W/280 W
32
56
16
64 (E5-2400)
64
8.2 TB
10.8 TB
6.14 TB
12.3 TB
12.3 TB
12
12
12
4
15
External SAN
External SAN Internal SAN
External SAN
External SAN
External SAN Internal distributed storage
6.144 Tbit/s
7.168 Tbit/s
1.28 Tbit/s
?
15.6 Tbit/s
120
64
16
?
128
4
2
1
3
5
Multi-plane networking
Y
Y
N
Y
Y
10GE/FCoE converged switch module
Y
Y
Y
Y
Y
40GE
Y
N
Y
Y
Y
RoCE
N
Y
N
N
Y
16G FC
N
Y
N
Y
N
PCIe expansion
Maximum number of standard PCIe slots per compute node
N
2 PCIe x8 + 2 x PCIe x16 slots
N
N
4 PCIe x8 or 2 PCIe x16 slots
Operating environment
Ambient temperature
5°C to 35°C
5°C to 40°C
5°C to 35°C
5°C to 35°C
5°C to 40°C
Engineering capability
Power supply and heat dissipation for front and rear slots
Computing capability
Computing capability per chassis
Memory capacity
Memory capacity per chassis (mainstream chips) 2.5" disk quantity per slot
Storage capability
Convergence of computing and storage resources backplane capability Number of 10GE ports per chassis Maximum plane quantity
Network switching
55
Chassis Specifications Comparison--E9000 vs. Tier 1 Competitive Products HUAWEI E9000
IBM Flexsystem
HP c7000
Cisco UCS 5108
Dell M1000e
Target field
HPC, virtualization, and high-end enterprise applications
Converged architecture
Large enterprise data center
Enterprise data center
Universal compute node
Dimensions
12U, 20.9" (530.2 mm) H 17.4" (442 mm) W 33.1" (840 mm) D
10U, 17.3" (440 mm) H 17.4" (442 mm) W 31.5" (800 mm) D
10U, 17.4" (442 mm) H 17.6" (447.04 mm) W 32.0" (813 mm) D
6U, 10.5" (267 mm) H 17.5" (445 mm) W 32" (812 mm) D
10U, 17.3" (440 mm) H 17.6" (447 mm) W 29.7" (754 mm) D
Maximum number of supported compute nodes
10U/16 FH or 8 FH server blades (vertical) blades (horizontal)
10U/14 FH or 7 FH server blades (vertical) blades (horizontal)
10U/16 FH or 8 FH server blades (vertical) blades (vertical)
10U/8 FH or 8 FH server blades (vertical) blades (horizontal)
10U/16 FH or 8 FH server blades (vertical) blades (vertical)
PSU
Max 6 x 3000 W AC or 6 x 2500 W DC
Max 6 x 2500 W AC
Up to 6 x 960 W/2250 W Output Up to 6 x 2400 W Output Single-phase AC or -48 V DC
4 x 2500 W 220 V single-phase AC
6 x 2700 W hot-plug PSUs 110 V to 120 V or 208 V to 240 V AC
Fan module
14 hot-swappable fan modules
10 hot-swappable fan modules
10 fan modules
8 hot-swappable fan modules
9 hot-swappable fan modules
Switch module
≤4 GE, 10GE, 8G FC, FCoE, and InfiniBand QDR/FDR
≤4 GE, 10GE (40GE ports available), 8G/16 G FC, FDR InfiniBand, SR-IOV, and FCoE
≤8 GE, 10 GE, 8 GFC, QDR InfiniBand, and SAS
≤2 Each switch module providing 4 x 10GE SFP+ ports (FCoE)
≤6 GE, 10GE, 8G FC, and InfiniBand QDR/FDR
Other features
Most hard disks supported on a compute node Support for DIMMs at 1.5 times the usual height
IBM new-generation blade server, planned for replacing the BladeCenter products
Widely deployed enterprise-level blade server released in 2007
Released in 2009 Important part of the UCS
Similar architecture with the HP c7000 64 server blades supported in four chassis in 40U
2-socket compute node
CH121/CH140 (twin node)/CH221/CH222
X86: X220/X222 (twin node)/X240 Power: p24L/p260
BL460c Gen8, BL420c Gen7, BL620c G7, BL465c Gen8…
B200 M2/B250 M2/B230 uiM2/B200M3/B22 M3
M420 (twin node), M520, M620…
4-socket compute node
CH240/CH242
X86:x440 Power:p460
BL680c G7 (4 blades), BL685c G7 (8 blades), and BL660c Gen8 (8 blades)
B440 M2/B420 M3
M820, M910, M905
Weight
Empty chassis: 70 kg (154.32 lb) Fully configured with 2-socket compute nodes: 212 kg (467.38 lb) Fully configured with 4-socket compute nodes: 223 kg (491.63 lb)
Empty chassis: 99 kg (218 lb) Fully configured: 224 kg (493 lb)
68.5 kg (DC PSUs, compute nodes only in the chassis) 87 kg (single-phase AC) 100.7 kg (three-phase AC)
Configured with compute nodes only in the chassis: 40.8 kg (89.95 lb) Full configuration: 115.7 kg (255.07 lb)
Max fully loaded with server blades and rear modules: 178 kg
56
Compute Node Comparison-E9000 vs. IBM Flexsystem CH121 Half-width 2-socket compute node
Flexsystem X222
CH140
E9000 Compute Node CH221 CH222
CH220
Half-width 2 x 2-socket twin compute nodes
Full-width I/O expansion node
Full-width I/O expansion node
CH240
Full-width storage expansion node
CH242
Full-width 4-socket E5-4600 compute node
Full-width 4-socket E7-4800 compute node
D Half-width 2 x 2-socket twin compute nodes
X240 A Half-width 2-socket compute node
X240 + I/O expansion node A
A
Full-width 2-socket + I/O expansion node
A
DIMMs at 1.5 times the usual height
G
More DIMMs supported
B
More standard PCIe cards supported
C
Larger local storage capacity
H
I
Others
Flexible configuration of RAID controller cards
57
D
CPUs of later series
E
Independent maintenance
F
Long-term stable operating at 40°C
Compute Node Comparison--E9000 vs. IBM Flexsystem CH121 Flexsystem X440
Half-width 2-socket compute node
CH140
E9000 Compute Node CH221 CH222
CH220
Half-width 2 x 2-socket twin compute nodes
Full-width I/O expansion Full-width I/O expansion node node
CH240
Full-width storage expansion node
CH242
Full-width 4-socket E54600 compute node
A/C/H
Full-width 4-socket E74800 compute node
B/C/D/H
Full-width 4-socket E5-4600 compute node
X240 + storage expansion node A/C Full-width 2-socket +storage expansion node
• •
The CPU with the highest performance in the Intel E5-2600 series is the E5-2690 CPU (20 MB cache, 8 cores, 2.90 GHz, and 135 W). The CPU with the highest performance in the Intel E7-4800 series is the E7-4870 CPU (30 MB cache, 10 cores, 2.40 GHz, and 130 W). A
DIMMs at 1.5 times the usual height
G
More DIMMs supported
B
More standard PCIe cards supported
C
Larger local storage capacity
H
I
Others
Flexible configuration of RAID controller cards
58
D
CPUs of later series
E
Independent maintenance
F
Long-term stable operating at 40°C
How to Defend Against IBM Flexsystem How to defend against the Flexsystem? •
The Flexsystem supports the internal storage device V7000. The V7000 occupies 3U space and has limited expansion capability. The Huawei external storage device S2600T or S5500T provides the same storage capacity as the V7000 in a just 2U space and higher expansion capability than the V7000.
•
A Flexsystem I/O expansion node provides more PCIe slots than the E9000 CH221. The Flexsystem I/O expansion node uses a PCIe switch to provide PCIe slots. However, the communication bandwidth between the switch and the CPU is the bottleneck. The E9000 CH221 provides two PCIe x16 slots directly from the CPU, supporting a higher
bandwidth. •
The Flexsystem provides 16G FC ports. The mainstream products provide 8G FC ports because the 16G FC specification has poor cost effectiveness.
•
The E9000 provides the 4-socket E7-4800 CH242 compute node, while IBM does not provide any similar product. Persuade customers to use the high-end E7-4800 compute node to beat IBM.
59
Compute Node Comparison--E9000 vs. HP C7000 CH121
CH140
Half-width 2-socket compute node
C7000 Compute Node
CH220
Half-width 2 x 2-socket twin compute nodes
Full-width I/O expansion node
E9000 Compute Node CH221 CH222
CH240
CH242
Full-width I/O expansion node Full-width storage expansion Full-width 4-socket E5-4600 compute node node
Full-width 4-socket E7-4800 compute node
BL460c Gen8
Half-height 2-socket E52600 compute node
A/B/F/G/I I: The 460c has to use four fewer DIMMs when E5-2690 series CPUs are configured.
BL660c Gen8 A/C/F/G
B/C/D/F
Full-height 4-socket E54600 compute node
BL680c Gen7 B/C/F/I I: The 680c occupies four slots. A chassis supports only four 4-socket E7 compute nodes.
Full-height 4-socket E74800 compute node
• •
The CPU with the highest performance in the Intel E5-2600 series is the E5-2690 CPU (20 MB cache, 8 cores, 2.90 GHz, and 135 W). The CPU with the highest performance in the Intel E7-4800 series is the E7-4870 CPU (30 MB cache, 10 cores, 2.40 GHz, and 130 W). A
DIMMs at 1.5 times the usual height
G
More DIMMs supported
B
More standard PCIe cards supported
C
Larger local storage capacity
H
I
Others
Flexible configuration of RAID controller cards
60
D
CPUs of later series
E
Independent maintenance
F
Long-term stable operating at 40°C
How to Defend Against the HP C7000 How to defend against the C7000? •
•
•
The BL680 supports 64 DIMMs. A chassis supports four BL680s, and each BL680 is configured with 64 DIMMs. This configuration provides low density and supports a complex stacking structure, which induces a high failure rate. The E9000 CH242 supports 32 DIMMs. A chassis supports eight CH242s, which provides higher density and supports more hard disks and PCIe cards than the BL680. GE + FC networking The E9000 currently does not support GE and FC networking. Persuade customers to use the 10GE and FC networking. If GE or FC networking is required, recommend the E6000 GE and FC, or E9000 10GE and FC networking to customers. Currently, the E9000 GE and FC solutions are under development and will be released in April of 2014. Brocade FC switch The E9000 that supports the Brocade FC switch was on sales in October of 2013.
61
Compute Node Comparison--E9000 vs. Cisco UCS CH121 UCS B200M3
Half-width 2-socket compute node
CH140
CH220
Half-width 2 x 2-socket twin compute nodes
Full-width I/O expansion node
E9000 Compute Node CH221 CH222 Full-width I/O expansion node
CH240
Full-width storage expansion node
CH242
Full-width 4-socket E5-4600 compute node
Full-width 4-socket E7-4800 compute node
A/B/F Half-width 2-socket compute node
B420M3 A/C/F Full-width 4-socket E54600 compute node
B440M2
C/B/F/I I: supports E7-4870 (130 W). The server blades that use the 130 W E74870 CPUs were recalled due to the overtemperature.
Full-width 4-socket E74800 compute node
• •
The CPU with the highest performance in the Intel E5-2600 series is the E5-2690 CPU (20 MB cache, 8 cores, 2.90 GHz, and 135 W). The CPU with the highest performance in the Intel E7-4800 series is the E7-4870 CPU (30 MB cache, 10 cores, 2.40 GHz, and 130 W). A
DIMMs at 1.5 times the usual height
G
More DIMMs supported
B
More standard PCIe cards supported
C
Larger local storage capacity
H
I
Others
Flexible configuration of RAID controller cards
62
D
CPUs of later series
E
Independent maintenance
F
Long-term stable operating at 40°C
How to Defend Against the Cisco UCS How to defend against the Cisco UCS? •
Storage or PCIe expansion computing nodes UCS does not provides blades such as storage expansion node(E9000 CH222) or PCIe expansion node(E9000 CH220/221 alike);
•
Open protocol and different network type UCS only supports 10GE FCoE, and the network environment is not open to regular equipment because of its private protocol, so it can not meet the demand of interconnection availability of blades and third party’s TOR switches, besides, it also can not compliant to the features of supporting different network type such as FC, GE, or physically isolated multi-plane networking;
63
Compute Node Comparison--E9000 vs. Dell M1000e CH121 M1000e M420
CH140
Half-width 2-socket compute node
Half-width 2 x 2-socket twin compute nodes
CH220 Full-width I/O expansion node
E9000 Compute Node CH221 CH222 Full-width I/O expansion node
CH240
CH242
Full-width 4-socket E54600 compute node
Full-width storage expansion node
Full-width 4-socket E74800 compute node
C/D/G 1/4-height, high-density node
M620
Half-height 2-socket compute node
M820
A/B/F/I I: The M620 cannot support the E5-2690 CPUs. The M620 supports four fewer DIMMs when the E52667 CPUs are configured.
A/C/F Full-height 4-socket E54600 compute node
M910 B/C/F/I I: The M910 supports the E7-4830 CPUs of up to 105 W.
Full-height 4-socket E74800 compute node A
DIMMs at 1.5 times the usual height
G
More DIMMs supported
B
More standard PCIe cards supported
C
Larger local storage capacity
H
I
Others
Flexible configuration of RAID controller cards
64
D
CPUs of later series
E
Independent maintenance
F
Long-term stable operating at 40°C
How to Defend Against the Dell M1000e How to defend against the Dell M1000e? •
Six I/O switch modules The E9000 provides four physical I/O switch slots to implement five switching planes. The Dell M1000e supports only three switching planes. The E9000 provides more ports on switch modules than the Dell M1000e.
•
GE + FC networking The E9000 currently does not support GE and FC networking. Persuade customers to use the 10GE and FC networking. If GE or FC
networking is required, recommend the E6000 GE and FC, or E9000 10GE and FC networking to customers. Currently, the E9000 GE and FC solutions are under development and will be released in April of 2014. •
Brocade FC switch The E9000 that supports the Brocade FC switch was on sales in October of 2013.
65
Compute Node Comparison--E9000 vs. IBM BladeCenter CH121 BladeCenter HS23
Full-height 2-socket compute node
• • •
CH140
Half-width 2-socket compute node
Half-width 2 x 2-socket twin compute nodes
E9000 Compute Node CH221 CH222
CH220 Full-width I/O expansion node
Full-width I/O expansion node
CH240
CH242
Full-width 4-socket E54600 compute node
Full-width storage expansion node
Full-width 4-socket E74800 compute node
A/B//F/G/I I: I: The HS23 does not support the E52690 CPUs but supports only the highcost VLP DIMMs.
The BladeCenter is gradually replaced by the Flexsystem. The CPU with the highest performance in the Intel E5-2600 series is the E5-2690 CPU (20 MB cache, 8 cores, 2.90 GHz, and 135 W). The CPU with the highest performance in the Intel E7-4800 series is the E7-4870 CPU (30 MB cache, 10 cores, 2.40 GHz, and 130 W).
A
DIMMs at 1.5 times the usual height
G
More DIMMs supported
B
More standard PCIe cards supported
C
Larger local storage capacity
H
I
Others
Flexible configuration of RAID controller cards
66
D
CPUs of later series
E
Independent maintenance
F
Long-term stable operating at 40°C
Agenda 1
Click toPositioning add Title Market Overview and
2
Click to add Product Specifications andTitle Highlights
3
Click to add Title Application Scenarios
4
Product Comparison
5
Ordering Guide
6
How to Get Resources 67
Quotation Elements for the Blade System Server Blade
Blade Chassis
集群NAS引擎
Blade Type
Type of the server blade
Chassis Type
Server blade
Base 存储单元 Parts
One server need base parts
Base Parts
Chassis need base parts
Mezz Parts
Mezz Card
Switch Module
Network switch
Optional parts on demands
Opt Parts
Optional parts on demands
服
Opt Parts
务
68
Blade System Product Configuration Procedure Start the UniSTAR tool. Select a model of Server blade.
Select base parts.
(optional)
Start the UniSTAR tool. Select a blade from the E6000 or E9000
Select a model of blade chassid.
CPU type
CPU quantity
Memory type
Memory quantity
Hard Disk type
Hard Disk quantity
Select the chassis according to blade
Management module quantity Select base parts.
Fan module quantity Power type
Power quantity
RAID card type Select switch. Select mezz card.
Select optional parts.
Select Fibre Channel modems.
Select network switch according to blade
Select network internet card Determine the PCIe card quantity.
Determine if the riser card need.
Select PCIe card type in the PCIe slots.
Determine if the OS need
Select OS type.
Select OS version
Determine if the DVD need
Determine if the insert card need
69
Select Fibre Channel modems.
Select optional parts.
Determine if the rail need
Determine if the DVD need
What Is SCT? Back to the Agenda
SCT: a swift, smart, and simple online configuration quotation tool SCT: simplifying configuration quotation
Quick learning about Huawei products
Intelligent product configuration
Product The SCT comprehensively shows
Configuration
the latest and hot Huawei products, updates data in real time, and enables quick product selection, helping you learn about Huawei products conveniently.
70
Intelligent product configuration, automatic configuration verification, and configuration sharing simplify product configuration, ensure configuration accuracy, and help you complete configuration easily.
Easy quotation order creation Quotation Easy quotation order creation, specific discount settings, automatic price verification, and convenient quotation order management help you comprehensively create and manage quotation orders.
How to Obtain SCT Access Permission Account/Permission application For a dealer
For a Huawei product manager
If you do not have a Huawei account, register an account at Huawei's official website. If you already have a Huawei account, send your account to the channel manager to apply for the SCT permission.
If you already have the eCFG permission, use your W3 account for login.
If you do not have the eCFG permission, use the permission application e-Flow to apply for data package permission of the corresponding product line.
SCT login Use your Huawei account to log in to the following website: http://app.huawei.com/unistar/sct
NOTE: You can log in to the SCT as user Guest to view and configure products. Enter the verification code to directly log in to the SCT. The recommended resolution is 1280 x 1024. Supported browsers include Internet Explorer 8.0, Firefox 11.0+, Safari 5.12+, and Chrome 19.0+. 71
SCT functions 1 Product Enables you to select a product for configuration.
2 Configuration Enables you to create a configuration item,
view the configuration list, add or delete a product, and modify product configuration.
3 Quotation Enables you to create a quotation item, view the quotation list, and add, delete, or modify product configuration.
72
Agenda 1
Click toPositioning add Title Market Overview and
2
Click to add Product Specifications andTitle Highlights
3
Click to add Title Application Scenarios
4
Product Comparison
5
Ordering Guide
6
How to Get Resources 73
How to Get Huawei Document Resource Method 1: enterprise website
http://enterprise.huawei.com/en/
Policy
– Learn about the channel partner policy. Query and Application
– Query or apply for partners. Documents and Tools
– Query pre-sales documents and tools of partners. Solutions
– Obtain solution cooperation information.
Method 3: Document User Guide
Method 2: email How can I obtain pre-sales documents?
[email protected] 74
How to Get Pre-sale Help
[email protected] Huawei Experts Team Partners
Support Center
7*24 hours presales email and
Partners
http://enterprise.huawei.com /en/about/contact
75
telephone support Products and solutions consulting
HUAWEI ENTERPRISE ICT SOLUTIONS A BETTER WAY
Copyright©2012 Huawei Technologies Co., Ltd. All Rights Reserved. The information in this document may contain predictive statements including, without limitation, statements regarding the future financial and operating results, future product portfolio, new technology, etc. There are a number of factors that could cause actual results and developments to differ materially from those expressed or implied in the predictive statements. Therefore, such information is provided for reference purpose only and constitutes neither an offer nor an acceptance. Huawei may change the information at any time without notice.
