Transcript
HPE StoreVirtual 3200 Storage best practices guide Guidelines for choosing and deploying the HPE StoreVirtual 3200 Storage system
Technical white paper
Technical white paper
Contents Executive summary ................................................................................................................................................................................................................................................................................................................................ 3 Introduction ...................................................................................................................................................................................................................................................................................................................................................3 Terminology.................................................................................................................................................................................................................................................................................................................................................. 3 General best practices ......................................................................................................................................................................................................................................................................................................................... 4 Building an array—where to start ...................................................................................................................................................................................................................................................................................... 4 Uninterruptible Power Supplies ...........................................................................................................................................................................................................................................................................................5 Network switching ............................................................................................................................................................................................................................................................................................................................5 Hardware setup best practices.................................................................................................................................................................................................................................................................................................... 5 Steps to configure a new HPE StoreVirtual 3200 ..................................................................................................................................................................................................................................................... 6 Creating RAID groups ................................................................................................................................................................................................................................................................................................................... 6 Network configuration ..................................................................................................................................................................................................................................................................................................................7 After the configuration........................................................................................................................................................................................................................................................................................................................ 9 Settings ....................................................................................................................................................................................................................................................................................................................................................... 9 Licensing ............................................................................................................................................................................................................................................................................................................................................... 10 Thin provisioning ................................................................................................................................................................................................................................................................................................................................. 10 Adaptive Optimization (Tiering) ............................................................................................................................................................................................................................................................................................ 12 Scaling out an HPE StoreVirtual 3200 management group........................................................................................................................................................................................................................ 13 Multi-Site Stretch Cluster considerations ...................................................................................................................................................................................................................................................................... 15 Performance considerations ...................................................................................................................................................................................................................................................................................................... 15 Scaling considerations for performance ................................................................................................................................................................................................................................................................... 15 Volume level performance factors ................................................................................................................................................................................................................................................................................. 15 Networking considerations for performance ........................................................................................................................................................................................................................................................ 16 Stay current on software and firmware ............................................................................................................................................................................................................................................................................17 Summary ...................................................................................................................................................................................................................................................................................................................................................... 18 For more information........................................................................................................................................................................................................................................................................................................................ 18
Technical white paper
Page 3
Executive summary This document is intended to highlight the best practices for the HPE StoreVirtual 3200. It is targeted toward a customer who is purchasing or setting up a new array or a customer who is seeking guidance on how to use the features of an array that is already up and running.
Introduction HPE StoreVirtual 3200 Storage is a storage array that scales from entry to mid-range enterprise applications. Like other HPE StoreVirtual storage products, HPE StoreVirtual 3200 is designed to fit into small rack spaces with a 2U rack profile for the base array enclosure that can be scaled up with up to three 2U drive enclosures. The management interface is hosted on the array so that no server-based utility is needed to configure or monitor the system. HPE StoreVirtual 3200 can also scale out by clustering multiple arrays together for higher performance, capacity, and data availability. HPE StoreVirtual 3200 carries forward familiar HPE StoreVirtual concepts such as Network RAID, Remote Copy (asynchronous volume replication), and Multi-Site Stretch Clusters (synchronous volume replication when combined with Network RAID). This paper discusses all of these features and shows ways to use them to achieve the best user experience with HPE StoreVirtual 3200 Storage.
Terminology Array enclosure—A chassis with two storage controllers, power supplies, fan modules, and a set of drives Drive enclosure—A chassis with two I/O modules, fan modules, power supplies, and a set of drives FC—Fibre Channel HDD—Hard disk drive I/O module—The circuit board that provides inter-enclosure SAS connectivity for shelves in an HPE StoreVirtual 3200 array LFF drive—Large form factor drive; an HDD containing platters that are approximately 3.5 inches (or 88.9 millimeters) wide; LFF SSDs are designed to fit in the same space as LFF HDDs LTU—License to use Management group—One or more storage pools that have a common management configuration NIC—Network interface card SAS—Serial attach SCSI, an interface standard used to connect host computers to peripheral devices such as solid state drives and disk drives Scale out—Add capacity and controllers to a management group; in terms of HPE StoreVirtual 3200, cluster two storage systems together in a single management group and combine their capacities to form a larger storage pool Scale up—Add capacity to a storage system; this can be done by adding drives to an enclosure and/or adding drive enclosures to the storage system SFF drive—Small form factor drive; an HDD containing platters that are approximately 2.5 inches (or 63.5 millimeters) wide; SFF SSDs are designed to fit in the same space as SFF HDDs SSD—Solid state drive Storage controller—The circuit board containing the storage computer with independent I/O paths to the storage network Storage pool—An aggregation of storage resources from one or more storage systems from which volumes can be provisioned Storage system—The hardware that consists of the array enclosure and all of its SAS-connected drive enclosures SVOS—StoreVirtual Operating System—the system software running on HPE StoreVirtual 3200 (formerly called LeftHand OS)
Technical white paper
Page 4
General best practices HPE StoreVirtual 3200 supports many different models of drives and each drive model is automatically placed into one of three tiers of storage by the storage system. SSDs are considered Tier 0; 10K RPM and 15K RPM SAS HDDs are considered Tier 1; and 7.2K RPM SAS drives are considered Tier 2. Table 1. Sample HPE StoreVirtual 3200 drive options—see QuickSpecs for current choices Drive type and form factor
Drive capacity (GB)
SFF SSD
200 400 800 1600 3200
LFF SSD
400 800
SFF 10K RPM SAS
300 600 900 1200 1800
SFF 15K RPM SAS
300 600
LFF 7.2K RPM MDL-SAS
2000 4000 6000 8000
SFF 7.2K RPM MDL-SAS
2000
When selecting drives for an array, avoid using different drive models within the same tier. In other words, don’t mix 10K RPM and 15K RPM SAS HDDs in an array—choose one type and one capacity to build the storage pool. Using the same drive model throughout will yield more predictable performance and reliability. Another factor to consider when building an array is that HPE StoreVirtual 3200 supports two tiers of storage in the array, even though there are three tiers defined. In most cases, build the array with enough Tier 1 storage to meet typical daily workload capacity and performance requirements. Then, add Tier 0 storage to boost performance in random I/O use cases or add Tier 2 storage to increase capacity for archive cases. HPE StoreVirtual 3200 Advanced Data Services LTU is required to enable intelligent movement of data between storage tiers, as described in the Adaptive Optimization section elsewhere in this paper.
Building an array—where to start Choosing the best starting point depends on the planned use model for the array. It also depends on whether or not there is a desire to implement a hybrid array with two tiers of storage. For example, for an array with a single tier of storage where performance is a high priority, 24 or 48 SFF 15K SAS drives would be a good starting point. For mixed workloads, consider starting with 24 or 48 SFF 10K SAS drives, where the capacity choices are much greater and the cost is much lower. If large capacity is a more important factor than performance, consider using 12 or 24 LFF 7.2K MDL-SAS drives.
Technical white paper
Page 5
For hybrid arrays, there are a couple of use cases—boosting performance or archiving data. To boost performance, add some SSDs to a quantity of SFF 10K SAS drives. The numbers depend on the capacity requirements but a few good models would be four SSDs with 12 10K SAS drives, six SSDs with 18 10K SAS drives, or eight SSDs with 40 10K SAS drives. It’s not generally advised to populate an HPE StoreVirtual 3200 array with more than 12 SSDs. Typically, the SSD capacity in an HPE StoreVirtual 3200 Storage system should range from five to 15% of the active (non-archive) capacity. For archive tiering, consider 24 SFF 10K SAS drives with 24 LFF 7.2K MDL-SAS drives.
Uninterruptible Power Supplies It is important to include a reliable power source when planning an HPE StoreVirtual 3200 Storage system deployment (or any server or storage device deployment). Hewlett Packard Enterprise supports using any online or double conversion UPS in conjunction with HPE StoreVirtual 3200. Avoid offline, standby, or line-interactive UPS implementations as these devices may not be able to switch over from the primary power source to the secondary power source fast enough to ensure continuous power to HPE StoreVirtual 3200.
Network switching Network interfaces on the HPE StoreVirtual 3200 can connect to industry-standard network switches (adhering to the IEEE 802.3 Ethernet standard). While any compliant Ethernet switch works, the capabilities and feature sets of the switch are key to designing a robust and resilient network for storage. Key design criteria for iSCSI networks should address: • Performance • Reliability • High availability • Low latency To address these design criteria, Hewlett Packard Enterprise recommends selecting switches that support the following software and hardware features: Software features • Jumbo frame support—See the jumbo frames discussion in Networking considerations for performance • Flow control—Also discussed in Networking considerations for performance • Loop detection—Spanning Tree Protocol (STP) helps ensure that only one active path exists between two or more interconnected switches • Quality of services • Link aggregation protocol • Traffic filtering (access controls) • Virtual local area networks (VLANs) Hardware features • Non-blocking backplane • Sufficient port buffering—allocate at least 512 KB of buffer cache for each port used for iSCSI • Wire-speed throughput
Hardware setup best practices For the most part, the hardware setup best practices are documented in the HPE StoreVirtual 3200 Storage System Installation Guide on the HPE StoreVirtual 3200 Storage documentation site referenced in the For more information section. Still, it is worth stressing a few items that should be addressed prior to configuring the storage system. • If the storage system is going to be scaled up with drive enclosures beyond the base array enclosure, make sure that the enclosures are all grouped together in the rack. The 0.5 m SAS cables included with the product are, in most cases, of sufficient length for this type of deployment. The exception is that for a storage system with three drive enclosures—in this case, one 1.0 m SAS cable is required to connect the third drive enclosure to the array enclosure.
Technical white paper
Page 6
• Within an array or drive enclosure, group drives of the same type are together in consecutive drive bays rather than interleaving different types of drives. For instance, if the storage system includes SSDs and 10K SAS HDDs, place all the SSDs together before putting any HDDs in the enclosure. This is mainly for organizational purposes and helps minimize latency on the SAS interface. Do not shuffle drive locations if adding capacity after deploying the array—once the initial RAID sets have been configured, the drives should not be moved. • In a hybrid storage system where both SSD and HDD technologies are used, install the SSDs in the array enclosure rather than a drive enclosure. This will reduce latency for the higher performing drives. • Fill the drive bays in an enclosure sequentially starting with the lowest-numbered drive bay. Don’t leave gaps between drives. Again, this is an organizational best practice.
Steps to configure a new HPE StoreVirtual 3200 The HPE StoreVirtual 3200 Storage System Installation Guide gives detailed instructions on how to install the HPE StoreVirtual 3200 hardware and perform the initial setup. It offers much more depth than HPE StoreVirtual 3200 Storage Quick Start Guide that ships in the box with the array. In addition to these documents, the array’s configuration setup process supplies extensive help on the right side of the browser window throughout the steps to set up the storage system. This section gives some additional insights into some of the configuration options.
Creating RAID groups During the initial setup of an HPE StoreVirtual 3200 Storage system, all available drives are displayed with the recommended RAID configuration preselected. All available drives are configured into RAID devices as part of the setup process, though each storage tier is configured separately. Depending on the number of drives present in the tier, different options might be presented for RAID configurations (for instance, a system with four drives present will have one option, but a system with 21 drives will have four options for RAID configurations). The resulting RAID devices will be split evenly between the two storage controllers in the HPE StoreVirtual 3200. Some things to keep in mind about the RAID configuration: • All the drives in the same storage tier must be configured in the same RAID level (RAID 5, RAID 6, or RAID 10). • Existing RAID devices cannot be expanded, but when new drives are added to the storage system, they can be configured as new RAID devices and added to the storage pool to increase the capacity of the storage system. – These new RAID devices must be the same RAID level as the existing RAID devices in the same storage tier. – Although it would be a best practice to do so, new RAID devices don’t necessarily have to include the same number of drives as the existing RAID devices. For instance, if the initial system configuration sets up the Tier 1 with five drives per RAID device, it is permitted to add more Tier 1 storage to the array with three or six drives per RAID device. – There is no rebalancing after a scale-up operation. When new data is written to the array, it is striped across all of the RAID devices in the storage pool (new and pre-existing), but the pre-existing data stays where it was unless it is moved as part of an Adaptive Optimization operation. This means that any performance benefit achieved by adding additional drives to a storage system will not apply to data that is already resident on the array. Table 2. HPE StoreVirtual 3200 Storage tiers Drive type
Tier
RAID mode
Minimum number of drives
Configuration
SSD
0
10
4
2 x (1+1)
5
6
2 x (2+1)
6
12
2 x (4+2)
10
4
2 x (1+1)
5
6
2 x (2+1)
6
12
2 x (4+2)
10
4
2 x (1+1)
6
12
2 x (4+2)
15K and 10K ENT SAS HDD
7.2K MDL-SAS HDD
1
2
Technical white paper
Page 7
• Each enclosure can contain (space permitting) up to one hot spare drive for each storage tier present in the array. Hot spares can help prevent data loss due to RAID failures by minimizing the amount of time between the failure of a drive and the start of rebuilding of the RAID device. Hot spares are not dedicated to any RAID device in the storage system—any RAID device in the same tier that contains drives of the same capacity or smaller can use them. When the array detects a failed drive or a predictive failure, it immediately begins to rebuild the RAID device using an available hot spare. When the failed drive is replaced, the array again rebuilds the RAID device with the new drive and returns the hot spare to its default state.
Figure 1. HPE StoreVirtual Management Console—configure RAID options for 21 drives
Network configuration As part of the initial configuration, the network ports on the storage controllers are set up in bonded configurations and supplied with IP addresses. For Fibre Channel models, the two 1GbE ports on each storage controller are bonded together and the bond is given a single IP address, which can be used to manage the storage system.
Technical white paper
Page 8
For 4-port iSCSI models, network ports 1G-3 and 1G-4 (for 1GbE storage controllers) or 10G-1 and 10G-2 (for 10GbE models) on each storage controller are bonded. This bond is given two IP addresses, which are used for iSCSI data traffic and management of the storage system. Two IP addresses are required for the system to optimize performance. Network ports 1G-1 and 1G-2, the onboard 1GbE ports, are disabled. In the case of the 8-port 1GbE iSCSI storage systems, network ports 1G-1, 1G-2, 1G-3, and 1G-4 on each controller are bonded together but two IP addresses are still used for the bond. The IP addresses for the two storage controllers must be on the same network and subnet, and must be valid for the network configuration of the environment into which HPE StoreVirtual 3200 is being deployed. Setting the default gateway is an optional step for the initial configuration but it is important to set this up some time to allow HPE StoreVirtual 3200 to communicate with hosts on remote networks. This is important for using HPE StoreVirtual 3200 features such as Remote Support and Software Updates, along with Multi-Site Stretch Cluster capability.
Figure 2. HPE StoreVirtual Management Console initial configuration of network settings
Technical white paper
Page 9
HPE StoreVirtual 3200 provides three options for the teaming/bonding of the network ports on each storage controller—Active-Passive, Link Aggregation Dynamic Mode (LACP), or Adaptive Load Balancing (ALB) bonding. The simplest solution, the Active-Passive bonding, utilizes only one of the available network ports per controller. The most complex solution, the LACP (802.3ad) bonding, delivers bandwidth aggregation and redundancy across both controller network ports but requires additional switch configurations such as IRF trunking. The preferred and recommended bonding mode is ALB, where the network ports may be connected to different switches, therefore, support switch failover. ALB bonding also offers the benefit of an increased transmission rate without additional switch configuration. If the network switches are set up as one big virtual switch (using features such as IRF trunking or Cisco vPC), the bonding on HPE StoreVirtual 3200 and servers connected to those switches must be set to LACP to avoid losing access to the storage in the event of a network component failure.
Important The network settings must be the same for the switches, clients, and storage systems. Set up the end-to-end network before creating storage volumes. After the bonds have been created on the network ports, the management of HPE StoreVirtual 3200 should be done using one of the IP addresses assigned to one of the bonds. The MGMT port used for the initial setup should be disconnected and only reconnected as directed by HPE Support or if you are configuring a separate network for management traffic.
After the configuration Settings Once the initial configuration is complete, navigate to the Settings page in HPE StoreVirtual Management Console and configure the proper values for DNS, email notification, and proxy settings. DNS settings help HPE StoreVirtual 3200 resolve host names used for Proxy, Remote Support, and Software Update functions. Email notification is a good way to monitor the health of the HPE StoreVirtual 3200 management group—at least one recipient should be designated to receive emails about critical- and warning-level events. Proxy settings allow HPE StoreVirtual 3200 Storage systems installed on private networks to access resources on public networks (again, this is useful for functionality such as Remote Support and Software Updates).
Figure 3. HPE StoreVirtual Management Console Settings page
Technical white paper
Page 10
Licensing If the Advanced Data Services Suite LTU has been purchased, it is a good practice to install the license on the array prior to creating volumes. It’s also a best practice to keep a copy of the license key in a safe place if it’s needed in the future (license keys are reapplied if the array is ever restored to factory default settings or the backplane is replaced). The HPE StoreVirtual Management Console User Guide provides information about obtaining and installing the Advanced Data Services Suite LTU.
Figure 4. HPE StoreVirtual Management Console Storage systems page—how to add a license key
Thin provisioning Thin provisioning, integrated throughout the HPE StoreVirtual architecture, allows storage for volumes, snapshots, remote copies, and volume clones to use only the storage they need. Thin provisioning helps to increase storage utilization and reduce total cost of ownership. The thin provisioning engine allocates storage from the HPE StoreVirtual 3200 storage pool as data is written to the volume (snapshot, Remote Copy, or clone) to be perpetually optimal for storage capacity utilization. All snapshots, remote copies, and clones are thin provisioned, eliminating the need for the administrator to manage the storage reservations for those items. Thin provisioning allows volumes to be sized large from the beginning with no impact on cost because no storage beyond a minimum 4 MB page is actually allocated when the volume is created. This approach, also known as overprovisioning, allows the administrator to present more capacity to the application servers than is actually in the storage pool. As storage volumes have blocks allocated to them and the storage pool begins to fill, the additional storage capacity can be added to the storage pool by scaling up and/or scaling out the HPE StoreVirtual 3200 Storage system. The storage pool capacity grows underneath those volumes with no impact to application availability.
Technical white paper
Page 11
If it is not possible to increase the capacity of a storage pool and there is a requirement to assure that space will exist for the volume, then it should be fully provisioned. Otherwise, Hewlett Packard Enterprise recommends thinly provisioning volumes on HPE StoreVirtual 3200 to make the most efficient use of the capacity. An administrator can change a volume from fully provisioned to thinly provisioned by simply editing the volume properties in the HPE StoreVirtual Management Console.
Figure 5. HPE StoreVirtual Management Console—edit volume properties dialogue
Conversely, a thinly provisioned volume can be changed to fully provisioned via the same method but only if sufficient capacity remains in the storage pool for the volume. HPE StoreVirtual Management Console will display a warning if an attempt is made to change a volume to full provisioned and there is not enough free capacity in the storage pool. Remember to consider Network RAID levels when calculating capacity requirements. The amount of space required is equal to the volume size multiplied by the Network RAID level factor (meaning the number of copies of data required by the Network RAID level—one for Network RAID 0, two for Network RAID 10, three for Network RAID 10+1, or four for Network RAID 10+2).
Technical white paper
Page 12
Adaptive Optimization (Tiering) Adaptive Optimization, or AO, is a feature that is licensed as part of the Advanced Data Services Suite. Each HPE StoreVirtual 3200 Storage system in a given storage pool must have an Advanced Data Services Suite LTU installed to enable Adaptive Optimization on the volumes that are built from the storage pool. AO moves data between storage tiers based on the access patterns for that data. AO is only useful if there are two tiers of storage (for example, Tier 0 SSDs and Tier 1 10K SAS HDDs) installed in the HPE StoreVirtual 3200 Storage system. When creating volumes on HPE StoreVirtual 3200, AO is one of the advanced options available. By default, Adaptive Optimization is “permitted” for a volume but the setting can be changed to “not permitted.” This is true whether the Advanced Data Services Suite LTU is installed and regardless of the number of tiers of storage present. But the setting has no effect unless the LTU is installed and more than one tier of storage is present in the storage pool.
Figure 6. HPE StoreVirtual Management Console volume creation dialogue
If two tiers of storage are present and the Advanced Data Services Suite LTU is not installed (or if AO is not permitted for a volume), all data written to a volume will go to the slower tier of storage until that tier is full, at which point data will start to be written to the faster tier. In this state, data will not be moved between tiers based on access patterns. If two tiers of storage are present and the Advanced Data Services Suite LTU is installed, all data written to a volume where AO is permitted will be stored in the faster tier of storage first until that tier is full, at which point data will start to be written to the slower tier. Also, HPE StoreVirtual 3200 will track data access patterns and move 4 MB pages of data between tiers based on how frequently the data is accessed.
Technical white paper
Page 13
Scaling out an HPE StoreVirtual 3200 management group Scale-out has been a feature in other HPE StoreVirtual products for many years. With HPE StoreVirtual OS version 13.5, it is possible to create a scaled-out HPE StoreVirtual 3200 configuration with two iSCSI storage systems in the same management group. Note that scale-out functionality is not currently supported with HPE StoreVirtual 3200 Storage FC systems. When scaling out an HPE StoreVirtual 3200 to cluster two storage systems together, a best practice is to use two identical storage systems. Use the same host interconnects (1GbE iSCSI, 10GbE iSCSI, or 10GBASE-T iSCSI) on both systems, use the same drive models and quantities, and use the same RAID settings for the drives. This will yield a predictable increase in performance, reliability, and storage capacity. If one storage system has a higher capacity than the other does, the resulting storage pool in the scaled-out system will be twice the capacity of the smaller system and the remaining storage in the larger storage system will be stranded. Typically, a scaled-out management group containing two 10GbE iSCSI storage systems each containing 24 10K RPM SAS HDDs will generally get higher performance than a single 10GbE iSCSI storage system with 48 10K RPM SAS HDDs. When ordering two HPE StoreVirtual 3200 Storage systems, at the same time, with the intention of deploying them in a scaled-out management group, it is fairly easy to ensure that the two systems have the same drive models and quantities but when upgrading an existing configuration it requires a bit of homework. To determine the drive information and RAID configuration of the currently deployed HPE StoreVirtual 3200, an administrator must log into HPE StoreVirtual Management Console, navigate to the Storage Systems page, and select Storage from the view selection list.
Figure 7. HPE StoreVirtual Management Console Storage Systems page showing view selection list
This opens the Storage view for the storage system. The RAID descriptions on this page provide information about the number of drives and RAID devices in each tier currently installed in the storage system.
Technical white paper
Page 14
Figure 8. HPE StoreVirtual Management Console—Storage Systems page RAID details
Navigating to the Disks page for the Storage System allows the administrator to verify the type and capacity of each drive in the storage system. Using this data, it’s possible to determine the type and quantity of drives to order for a new HPE StoreVirtual 3200 when scaling out a management group. In addition, the information about the RAID configuration can be used during the initial setup of the new storage system to ensure that it is configured the same way as the already-deployed storage system.
Figure 9. HPE StoreVirtual Management Console—Disks page
Technical white paper
Page 15
When scaling out from a single storage system to the next storage system, be sure to install the scale-out system using the Configuration Setup, stopping before the management group and storage pool creation step. You should then use HPE StoreVirtual Management Console to discover the newly installed scale-out system and add it to the management group and storage pool. When adding storage to a scaled-out system, be sure to add storage to each storage system to avoid stranded storage.
Multi-Site Stretch Cluster considerations In addition to the suggestions for scale-out, there are some best practices in configuring Multi-Site or stretch clusters on HPE StoreVirtual 3200 Storage system. In order to configure Multi-Site, you must install the Advanced Data Services Suite LTU on each storage system. Additionally, you need to install a Failover Manager at a third site. Separate subnets and power sources are recommended for each storage system. The recommended steps to configure Multi-Site are: • Set up a scaled out management group • Install the Failover Manager on the iSCSI network • Set up the sites from HPE StoreVirtual Management Console, designating only one site as the primary site • Specify the storage system and one or more servers for each site • Create a third site for the Failover Manager • Confirm that proper routing exists to allow the two storage systems and the failover manager to communicate across the iSCSI network Note that there are latency requirements for the network between the sites and the Failover Manager as described in the HPE StoreVirtual Management Console User Guide.
Performance considerations Scaling considerations for performance One of the most obvious ways to increase performance with HPE StoreVirtual 3200 is to scale out the management group from a single storage system to two storage systems. This adds processor and memory resources along with capacity. Another potential way to increase performance with HPE StoreVirtual 3200 is to scale up the storage system with more drives. As mentioned in Creating RAID groups, the data already resident on the storage system is not restriped across the new disk drives, so any performance improvement is only evident on new data or if existing data is copied to new volumes. In a pure Tier 1 storage system, adding more disk drives typically yields substantially more performance. Adding tier 0 storage to a Tier 1 system (and installing the Advanced Data Services Suite LTU) typically provides much better performance than the storage system is seeing without tier 0 storage. Depending on the use model, this may or may not be perceivable in production operating environments (as opposed to performance test scenarios). HPE StoreVirtual 3200 Storage system utilizes a multi-core processor. For optimal storage performance, configure at least four RAID devices per system. This allows the RAID engine to leverage additional processor resources. For example, given 12 10K HDDs, the system recommends (2) RAID 5 devices (5+1) as capacity optimized, however, for performance optimization there is a (4) RAID 5 device (2+1) option.
Volume level performance factors In most cases, Hewlett Packard Enterprise recommends using Network RAID 0 for volumes on a single HPE StoreVirtual 3200 Storage system for the best balance of volume availability, capacity utilization, and performance. With Network RAID 10, all the data is mirrored, so all data written to HPE StoreVirtual 3200 will require twice the number of write operations on the part of the storage controllers, and the data consumes twice as much capacity. With the dual controllers in HPE StoreVirtual 3200 Storage system, there’s no need to use Network RAID 10 on volumes to protect against a controller failure. Each controller has access to all of the drive RAID sets, so data availability is not impacted by using the more capacity-efficient Network RAID 0. One place where Network RAID 10 excels on a single HPE StoreVirtual 3200 Storage system is for read-intensive tasks—Network RAID 10 volumes typically deliver an excellent read performance. This is because the storage system has a copy of the data resident on the RAID devices managed by each of the two storage controllers.
Technical white paper
Page 16
Configure at least four volumes per storage system. HPE StoreVirtual 3200 Storage system supports active/active asymmetric LUN presentation, which means that all controllers actively service host I/O, but each LUN is serviced through a single port. Both storage controllers in an HPE StoreVirtual 3200 Storage system will be active when a minimum of two volumes are defined and exported to servers. The storage system also employs two target engines per controller to utilize processor resources better. Therefore, for optimal performance, a minimum of four volumes is required. If desired, choosing a total number of volumes in increments of four provides an even distribution of workload.
Networking considerations for performance Use jumbo frames for situations where the typical workload has large amounts of sequential data that needs to be transferred. When using jumbo frames, all devices on the same network segment (HPE StoreVirtual 3200, network switches, and server NICs) must have the same frame size. When jumbo frames are employed in iSCSI networks, remember to set the maximum transmission unit (MTU) on initiator access, trunk, and bonds for all appropriate ports that are within the iSCSI data path. HPE StoreVirtual 3200 won’t run faster but jumbo frames cut down on some of the overhead in the network protocols used to transfer data since more data can be moved with each transaction. HPE StoreVirtual OS version 13.5 is the minimum software version required to use jumbo frames with the StoreVirtual 3200.
Important The network settings must be the same for the switches, clients, and storage systems. Set up the end-to-end network before creating storage volumes. The default MTU in Ethernet is 1,500 bytes; the resulting frame with Ethernet header and preamble is 1,518 bytes. Any Ethernet frame larger than the default is considered a jumbo frame. The maximum allowed frame size supported by HPE StoreVirtual 3200 is 9,000 bytes. The setting in the HPE StoreVirtual 3200 Management Console is called Frame Size and can be found by editing the network bond properties on the Storage Controllers page. It is important to edit the bond on each storage controller.
Figure 10. Editing the Frame Size in HPE StoreVirtual Management Console
Technical white paper
Page 17
IP storage networks are unique for sustained bandwidth that is required to maintain adequate performance levels under heavy workloads. When a network port becomes saturated, excess frames may be dropped because the port can’t physically handle the amount of traffic it is receiving. Dropped frames then cause the packets to be re-sent and the overhead of re-sending the packets can add to the decreased performance. Ethernet Pause (Flow Control) technology can eliminate this problem by controlling the speed at which data is sent to the port. IEEE 802.1bb Priority Based Flow Control enhances a converged iSCSI/LAN network by pausing storage traffic that should remain lossless and allows the converged LAN traffic to buffer, queue, pass, or drop. Flow control ensures that storage packets don’t get dropped while LAN traffic still follows the inherent nature expected of TCP-based traffic. Hewlett Packard Enterprise recommends enabling Flow Control on all switch ports that are used for the iSCSI SAN to achieve optimal performance (including the LACP trunk ports of the inter-switch connection). There is no need to change any settings on the HPE StoreVirtual 3200 side—the network ports on the storage controllers automatically negotiate to match the flow control settings of the switch ports.
Stay current on software and firmware The HPE StoreVirtual 3200 dashboard shows the availability of any outstanding patches or system software version updates. These patches and updates are issued to fix problems or introduce new features. What’s more, Hewlett Packard Enterprise recommends that they are installed in a timely manner.
Figure 11. HPE StoreVirtual Management Console Dashboard showing available updates
Click on Updates Available or use the HPE StoreVirtual Management Console main menu to navigate to the Updates page to review and install patches and software updates. For more details, or for instructions on other methods of installing software updates, refer to the HPE StoreVirtual OS Update Guide on the HPE StoreVirtual 3200 Storage documentation site referenced in the For more information section. Remember, the Software Update functionality does not work properly unless the DNS, proxy, and gateway settings are properly configured. HPE StoreVirtual 3200 updates page also updates client software components such as Application Aware Snapshot Manager, the Failover Manager, and the DSM installer. The client components are upgraded separately from the HPE StoreVirtual OS and firmware. In addition, one can receive support alerts announcing product support communications, driver updates, software releases, firmware updates, and customer replaceable component information in their email by signing up at: hpe.com/support/e-updates.
Technical white paper
Summary HPE StoreVirtual 3200 offers a number of features. As a supplement to the product documentation, this white paper offers advice on choosing, deploying, and upgrading an HPE StoreVirtual 3200. There are tips for scaling up, scaling out, and modifying settings to achieve better performance. While no guide can cover every possible environment, the information in this document should prove helpful in most situations. Always consider the following steps when deploying a new HPE StoreVirtual 3200: • Use the HPE StoreVirtual Management Console Configuration Setup to configure an HPE StoreVirtual 3200 rather than trying to circumvent the process. • Enable Remote Support as part of the initial configuration and test connectivity after configuring DNS and proxy settings. • Disconnect from the MGMT port after completing the Configuration Setup. • Configure DNS, default gateway, and proxy after completing the Configuration Setup and prior to using the storage system in a production environment. • Turn on email notifications so that any alerts from the storage system are reported to the appropriate resource. • Validate that there are zero active alarms prior to deploying an HPE StoreVirtual 3200 into a production environment. • Check HPE StoreVirtual Management Console for HPE StoreVirtual OS software updates and patches periodically.
For more information HPE resources HPE StoreVirtual 3200 Storage HPE StoreVirtual 3200 Storage documentation HPE SAN Design Guide
Learn more at hpe.com/storage/storevirtual3000
Sign up for updates
© Copyright 2017 Hewlett Packard Enterprise Development LP. The information contained herein is subject to change without notice. The only warranties for Hewlett Packard Enterprise products and services are set forth in the express warranty statements accompanying such products and services. Nothing herein should be construed as constituting an additional warranty. Hewlett Packard Enterprise shall not be liable for technical or editorial errors or omissions contained herein. a00009194ENW, May 2017
