HPE StoreVirtual VSA Ready Nodes Reference configurations
Technical white paper
Technical white paper
Contents Introduction to software-defined storage and HPE StoreVirtual......... .......... ......... .......... ......... .......... ......... .......... ......... .......... ......... .......... ......... .......... ......... .......... ......... .......... ......... ......... .......... ......... 3 . Architectu Architectural ral overview overview................................. ................................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................... 4 ... Moving Moving data services services into the server server platform platform ............................... ............................................... ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ .................... 4 .... Essential Essential data services services for software-de software-defined fined storage storage................................. ................................................. ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. .................. 5 .. Integration Integration into hypervisor hypervisor ecosystem ecosystem ................................ ................................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................. ................................. ..................... 6 ..... Introducing Introducing HPE StoreVirtu StoreVirtual al VSA Ready Nodes................................. ................................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................. 6 . Reference Reference configurat configurations ions ............................... ................................................ ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ...................... 6 ...... Performance Performance estimates estimates ............................... ................................................ ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................. ................................. ........................... 7 ........... Networking Networking configurat configuration ion ................................. ................................................. ................................. ................................. ................................ ................................. ................................. ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. .................... 8 ... Storage Storage configurat configuration ion................................. ................................................. ................................ ................................. ................................. ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ .......................... ..........10 HPE StoreVirtua StoreVirtuall VSA Ready Node—Small Node—Small................................. ................................................. ................................ ................................. ................................. ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ..................... ..... 11 HPE StoreVirtua StoreVirtuall VSA Ready Node—Mediu Node—Medium m............................... ............................................... ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................12 HPE StoreVirtua StoreVirtuall VSA Ready Node—Mediu Node—Medium m Hybrid ............................... ................................................ ................................. ................................ ................................. ................................. ................................. ................................. ................................ ................................. ................................. ................................ ............................ ............13 HPE StoreVirtua StoreVirtuall VSA Ready Node—Large Node—Large Hybrid ................................ ................................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................. ................................. ................................ ................................. .................14 Summary Summary ................................ ................................................ ................................. ................................. ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................ ................15 Appendix Appendix A ............................... ................................................ ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ........................... ...........15 I/O Performance Testing Methodology with Vdbench ............................................................................................................................................................................................................................... 15 Workload Modeling Considerations................ Considerations................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ......................... ......... 15 Medium-Hybrid Configuration................. Configuration................................. ................................ ................................. ................................. ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ........................... ...........17 Medium Configuration Configuration................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ............................... ...............18 Vdbench Vdbench Test Plan and Definition Definitionss ................................. ................................................. ................................. ................................. ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. .............................. ..............19 Appendix Appendix B ............................... ................................................ ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ........................... ...........21 Baseline Baseline Performance Performance Results Results ................................ ................................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................. ................................. ................................ ................................. ................................. ................................ ................................. ...................... ..... 21 Reference Reference and additional additional resources resources ................................ ................................................ ................................. ................................. ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. ................................ ................................. ................................. .............................. ..............25
Technical white paper
Page 3
Introduction to software-defined storage and HPE StoreVirtual Shared and highly available storage i s at the center of every server virtualization project, as well as, a lot of client virtualization projects. Custom software-defined storage solutions present a third option delivering storage to these projects, sitting b etween traditional IT architectures with dedicated storage arrays, like HPE 3PAR StoreServ storage, and purpose-built virtualization solutions, like HPE Hyper Converged systems. In contrast, software-defined storage moves data services, previously provided by external storage arrays, into the server with maintained full control over the storage layer. HPE StoreVirtual VSA enables customers to run enterprise-class storage features on the same set of hardware that also runs the application workload, whether it’s virtualized da tabases or file services, for example. By running data services on industry-standard x86 servers side by side with virtualized application workloads to what the IT ind ustry regards as hyperconverged, customers can significantly reduce the overall solution componentry and its data center footprint, and lower the initial cost of deployment and of ongoing maintenance. Small- and medium-sized businesses as well as enterprise remote offices can benefit the most from this third deployment option.
HPE storage and servers
Software-defined environment
Storage array
Virtual machines
Virtual machines
With HDDs, SSDs
HPE ProLiant with HDDs and SSDs
HPE Hyper Converged with predefined storage blocks
Virtual machines
Virtual machines
HPE ProLiant with HDDs and SSDs
HPE Hyper Converged with predefined storage blocks
Server
Server
Dedicated storage fabric
HPE Hyper Converged
Converged network
Converged network
Figure 1. Traditional architectures deploy compute and storage resources separately; software-defined storage converges storage and compute.
One of the benefits of software-defined storage, having the flexibility of the underlying server platform, can also be a challenging aspect with the vast variety of server platforms and server options. This white paper outlines recommended reference configurations for HPE StoreVirtual VSA on the HPE ProLiant server platform, known as HPE StoreVirtual VSA Ready Nodes. This paper aims to a ccelerate the overall deployment time and simplify the installation of software-defined storage solutions with the HPE platform by presenting details of server configuration, software configuration, and the performance characteristics to be expected. Based on the information provided, system architects or administrators can build converged compute and storage for virtualization projects guided by a configuration tested by HPE. Audience for this paper A technical audience will gather an understanding of the HPE StoreVirtual Architecture, details of the reference configurations for HPE StoreVirtual VSA based on HPE ProLiant platforms, and the deployment of the reference configuration.
Important This paper assumes familiarity with the concepts and the deployment of virtualization solutions. Contact your HPE partner for more information on how to order and install the configuration presented in this paper.
Technical white paper
Page 4
Architectural overview The heritage of the HPE StoreVirtual VSA software is the scale-out iSCSI StoreVirtual storage array technology. In that scale-out storage architecture, the storage pool can start out small and grow storage capa city and performance by adding more storage systems to the pool as illustrated in Figure 2. Each storage s ystem contributes capacity and performance of its storage resources to the overall storage pool. This mode of scalability by building blocks delivers predictable growth and simplicity.
Figure 2. Scale-out storage delivers linear scalability of storage capacity and performance.
HPE StoreVirtual VSA also allows storage administrators to manage multiple s torage pools. Storage pools are comprised of instances of HPE StoreVirtual VSA with homogenous storage configurations. Different storage pools within the environment can be used to accommodate different performance, capacity, and data protection preferences. HPE StoreVirtual allows moving these workloads nondisruptively between storage pools using the built-in HPE Peer Motion technology. Among the reference configurations presented in this paper, you will find configurations that expose different compute and storage characteristics for different use cases. These reference configurations can be used as the basis for custom configurations that can be created to deliver unique capacity and performance profiles.
Moving data services into the server platform There are many environments where there are substantial benefits to moving the data services into the server. That is particularly true for applications with a fixed compute and storage ratio. For example, when adding more users to a given application, software-defined storage allows adding more compute and more storage in a single server building block. To converge compute and storage into a single physical server, HPE StoreVirtual VSA is provided as a utility virtual machine (VM) that sits on top of VMware vSphere®, Microsoft® Hyper-V or Linux® KVM, and virtualizes the s torage resources inside the server. It is visible as a regular virtual machine that is visible to the administrator, but it is a special VM that presents a block storage device (a disk resource via the iSCSI protocol) back to the hypervisor. It uses the network r esources supported by the hypervisor, such as 1GbE, 10GbE, and 25GbE Ethernet infrastructure, to communicate with the hosts accessing the storage resources provided by the HPE StoreVirtual VSA (via iS CSI) as well as other HPE StoreVirtual VSA instances in the HPE StoreVirtual storage pool. The virtualized storage resources can be any RAID-protected block storage option supported by the hypervisor: solid-state drives (SSDs), conventional hard disk drives (HDDs), or a mix of SSDs and HDDs with HPE StoreVirtual Adaptive Optimization technology. This allows HPE StoreVirtual storage pools to deliver unique performance characteristics, ranging from high-performing all-flash performance with low latency by using only SSDs to large capa city archive type storage pools with the use of large capacity drives. Each HPE StoreVirtual VSA instance requires a license for the storage capacity it is virtualizing on the host it is running on.
Technical white paper
Page 5
Storage pool
Virtual machines
Server With HDDs, SSDs
Converged network
Converged network
Figure 3. HPE StoreVirtual VSA virtualizes storage resources.
Figure 4. HPE StoreVirtual VSA pools storage resources of individual servers.
In Figure 3, a si ngle HPE StoreVirtual VSA instance is shown running on top of a server using a supported hypervisor with the resources used by it: storage and networking. HPE StoreVirtual VSA typically has exclusive access to the built-in storage resources in the server and shared access to the networking resources. The instance of the HPE StoreVirtual VSA represents a single storage system in the scale-out architecture. To pool the capacity of multiple servers and to provide high availability, two or more are required to be combined in a so-called storage pool, or storage cluster. Figure 4 shows three servers, all running HPE StoreVirtual VSA, on a single converged network that is used for applications and storage traffic. Applications running on the three hosts can store data on the storage pool. In most cases, there are two clusters in these environments: a hypervisor cluster like a vSphere Cluster or a Windows® Failover Cluster and a storage pool that presents storage back to these hosts. While this paper focuses on the converged deployment of HPE S toreVirtual VSA and application workloads on the same physical server, HPE StoreVirtual VSA can also be deployed on a set of dedicated servers in environments that require separation of storage and compute resources. In those cases, customers can achieve commonality of parts between the compute and storage systems.
Note For large environments with many different applications, a dedicated storage system allows for fine-grained control over storage resources independently of compute resources. This is especially true for applications with vastly different compute (CPU and memory resources) to storage (capacity and performance) ratios. Contact your HPE presales or HPE partner about the best architectural approach.
Essential data services for software-defined storage There are some critical data services that have traditionally been d elivered on shared storage arrays, which now must be accommodated by software-defined storage: scalability, high availability, and data mobility. For more in-depth information on the data services provided by HPE StoreVirtual products, please see the HPE StoreVirtual Architecture white paper and the HPE StoreVirtual product documentation.
Technical white paper
Page 6
Integration into hypervisor ecosystem Especially when converging compute and storage into the same physical server, virtualization administrators are covering server and storage administration. To simplify the workflow for the virtualization administrator, integration of software-defined storage management into the virtualization management software becomes critical. Hewlett Packard Enterprise provides management integration for VMware® vCenter™ and Microsoft System Center to allow typical management operations to be performed directly from the hypervisor management software. Storage capacity monitoring and storage provisioning are accessible for the virtualization administrators through HPE OneView for vCenter and HPE OneView for System Center. These components connect the hypervisor management software and the storage systems (all primary storage array families in the HPE portfolio). HPE OneView for vCenter is provided as a virtual appliance and HPE OneView for System Center is software that is installed on the System Center server(s). Hewlett Packard Enterprise recommends these management integrations for all HPE StoreVirtual VSA installations. The primary management console for HPE StoreVirtual VSA is the S toreVirtual Centralized Management Console (CMC), which allows complete control all aspects of the HPE StoreVirtual VSA systems, s torage pools, and storage volumes.
Note HPE storage integrations into Microsoft System Center and VMware vCenter are included in the price of the HPE StoreVirtual VSA product and do not require an additional license. HPE ProLiant Server Management functionality requires an HPE OneView license, which is included in the StoreVirtual VSA Ready Node reference configurations. HPE StoreFront Analytics for vRealize Suite provides further management and analysis features and requires an additional license.
Introducing HPE StoreVirtual VSA Ready Nodes By running on top of vSphere, Hyper-V, or KVM, HPE StoreVirtual VSA can add data services on virtually any industry-standard x86 server platform. This flexibility provided by the hypervisor and software-defined storage controller software allows system architects to choose from an even greater number of options on the selected server platform. HPE StoreVirtual VSA Ready Nodes are reference configurations to provide customers and partners a starting point when selecting configurations for virtualization projects on the HPE ProLiant server platform. HPE StoreVirtual VSA Ready Nodes will be r efreshed over time as new generations of servers and new options b ecome available.
Important This edition of the paper focuses on HPE StoreVirtual VSA Ready Nodes on Windows Server® 2012 R2 and VMware vSphere 6.x only. Partners and customers can make any supported changes to the hardware and software configuration to achieve different objectives.
Reference configurations
The reference configurations presented in this white paper are based on internally tested configurations of HPE ProLiant servers and HPE StoreVirtual VSA. They represent a well-known starting point for custom configurations for software-defined storage projects. The use cases that can be covered range from small branch office virtualization solutions to larger capacity solutions for data centers. Selecting a reference configuration from Table 1 requires a degree of understanding of the workload that is going to be virtualized. For technology refresh projects, it helps to monitor and metric the virtualized workload on an existing storage system or use the storage performance counters on the hypervisor to understand the base-level target for any new deployments. To calculate the usable capacity by a storage pool based on HPE StoreVirtual VSA Ready Nodes, multiply the number of systems by the capacity per system and divide the product by two to account for Network RAID 10 to deliver high availability.
Technical white paper
Page 7
Table 1. HPE StoreVirtual VSA Ready Nodes at a glance Configuration names
HPE StoreVirtual VSA Ready Node Small
HPE StoreVirtual VSA Ready Node Medium
HPE StoreVirtual VSA Ready Node Medium Hybrid
HPE StoreVirtual VSA Ready Node Large Hybrid
Primary use cases
Small and medium businesses; remote and connected offices
Virtualization in large remote and connected offices
Performance server virtualization projects
High-density server virtualization projects
HPE base server model
HPE ProLiant DL360 with 8 SFF
HPE ProLiant DL380 with 16 SFF
HPE ProLiant DL380 with 24 SFF
HPE ProLiant DL380 with 24 SFF
Processor
(2) Intel Xeon E5-2620 v4 with 10 cores @ 2.2 GHz
(2) Intel Xeon E5-2640 v4 with 10 cores @ 2.4 GHz
(2) Intel Xeon E5-2650 v4 with 12 cores @ 2.2 GHz
(2) Intel Xeon E5-2680 v4 with 14 cores @ 2.4 GHz
Memory
192 GB
256 GB
256 GB
512 GB
Networking
(8) 1GbE ports
(2) 10GbE ports + (4) 1GbE ports
(2) 10GbE ports + (4) 1GbE ports
(4) 10GbE ports + (4) 1GbE ports
Storage controller(s)
HPE Smart Array P440ar with 2 GB Flash Backed Write Cache
HPE Smart Array P840ar with 2 GB Flash Backed Write Cache
HPE Smart Array P440ar with 2 GB Flash Backed Write Cache, HPE Smart Array P840 with 4 GB Flash Backed Write Cache
HPE Smart Array P440ar with 2 GB Flash Backed Write Cache, HPE Smart Array P840 with 4 GB Flash Backed Write Cache
Hard disk drives
(5) 1.2 TB SAS HDD
(12) 1.2 TB SAS HDD
(7) 1.2 TB SAS HDD
(18) 1.2 TB SAS HDD
Solid-state drives
None
None
(3) 400 GB SAS SSDs
(6) 800 GB SAS SSD
Usable capacity per system
4 TB
10 TB
7.5 TB (10% flash)
18.5 TB (18% flash)
Recommended HPE StoreVirtual license
HPE StoreVirtual VSA 4 TB (sold as pack of 3 licenses)
HPE StoreVirtual VSA 10 TB
HPE StoreVirtual VSA 10 TB
HPE StoreVirtual VSA 50 TB
®
®
Note For quoting and purchasing purposes, partners and HPE presales can download the templates for Sal esBuilder for Windows on the Software-defined Storage section on the Storage Partner Portal and internal Worldwide Storage Sales Portal respectively.
Performance estimates The performance characteristics are based on engineering estimates for the HPE StoreVirtual VSA Ready Node on top of VMware vSphere. The workloads in Table 2 can be used to approximate performance of databases and other highly randomized workloads like virtualization. The numbers can be compared to generic storage performance numbers published by other storage vendors.
Important The performance numbers in Table 2 are estimates based on HPE internal performance testing and expert opinion.
Table 2. Estimated HPE StoreVirtual VSA Ready Nodes pe rformance Workload
HPE StoreVirtual VSA Ready Node Small
HPE StoreVirtual VSA Ready Node Medium
HPE StoreVirtual VSA Ready Node Medium Hybrid
HPE StoreVirtual VSA Ready Node Large Hybrid
8 KB random, 100% reads
550 IOPS @ 20 ms
1600 IOPS @ 20 ms
10,500 IOPS @ 5 ms (SSD) 950 IOPS @ 20 ms (HDD)
20,000 IOPS @ 5 ms (SSD) 2500 IOPS @ 20 ms (HDD)
8 KB random, 60% reads
350 IOPS @ 20 ms
1100 IOPS @ 20 ms
4200 IOPS @ 5 ms (SSD) 650 IOPS @ 20 ms (HDD)
8500 IOPS @ 5 ms (SSD) 1300 IOPS @ 20 ms (HDD)
8 KB random, 20% reads
300 IOPS @ 20 ms
900 IOPS @ 20 ms
2600 IOPS @ 5 ms (SSD) 500 IOPS @ 20 ms (HDD)
5200 IOPS @ 5 ms (SSD) 900 IOPS @ 20 ms (HDD)
8 KB random, 0% reads
400 IOPS @ 20 ms
1300 IOPS @ 20 ms
2200 IOPS @ 5 ms (SSD) 750 IOPS @ 20 ms (HDD)
4500 IOPS @ 5 ms (SSD) 1300 IOPS @ 20 ms (HDD)
Technical white paper
Page 8
These numbers are estimates for per-node performance in a three-node storage pool. To roughly size for a four- or six-node storage pool, multiply the per-node number by a factor of four or six respectively. Storage performance numbers can vary s ignificantly by workload. To achieve the numbers in Table 2, the workload generator needs to drive sufficient and consistent I/O to the configuration under test, and networking components may not introduce bottlenecks.
Networking configuration All HPE StoreVirtual VSA Ready Nodes use HPE ProLiant DL300 series onboard network ports and ports on FlexLOM cards (see Figure 5). This provides all configurations with resilient and high-bandwidth network connectivity and represents a good starting configuration for most virtualization projects. For all HPE StoreVirtual VSA Ready Node configurations, the HPE Aruba 2900 (especially for Small), HPE FlexFabric 5700 and HPE 5900 Switch Series (Medium, Medium Hybrid, and Large Hybrid) are recommended.
Figure 5. Network ports on HPE StoreVirtual VSA Ready Nodes (Small on the left; Medium, Medium Hybrid on the right; Large Hybrid not shown).
There are at least three network functions in every virtualization project using a software-defined implementation: management, storage, and virtual machine networking. In most environments, there will be additional virtual machine networks for further segmentation and isolation of certain network traffic. Ideally, these network functions are supported by res ilient network connectivity and do not compete for bandwidth to ensure stable operation. As shown in Figures 6 and 7, there are sufficient network ports to team the network ports for all network functions. Network teams provide resilience and, in some cases, increased bandwidth through bandwidth aggregation. For optimal resiliency, these network teams are connected over two separate network switches. In some configurations, it is also possible to configure network teams across two network cards for additional resiliency on a server component level.
HPE StoreVirtual VSA Ready Nodes
Network switches (1GbE)
1GbE port (onboard)
1GbE port (onFlexLOM)
Figure 6. Network ports on HPE StoreVirtual VSA Ready Nodes (Small).
Technical white paper
Page 9
HPE S toreVirtual VSA Ready Nodes
Network switches (1GbE)
Network switches (10GbE)
1GbE port (onboard)
10GbE port (on FlexLOM)
Figure 7. Network ports on HPE StoreVirtual VSA Ready Nodes (Medium and Medium Hybrid).
In addition to network teaming, it is recommended to separate the networking functions using virtual LANs (VLANs). For a typical HPE StoreVirtual VSA Ready Node deployment with the network functions mentioned previously, there are VLANs to separate each of the network functions. Table 3 outlines the number VLANs and IP addresses that need to be planned for including the subtle difference between the hypervisors. For example, Microsoft Windows Servers will need an extra IP resource for Windows Failover Clustering, and VMware vSphere will use two IP addresses for storage multipathing. Table 3. Recommended VLANs and IP planning VLAN
Network function
Description and IP planning
1
VM network
This is the primary network segment for virtual machines on the systems; in some deployments, more than one VLAN may be desired for further separation of the virtualized workloads. HPE StoreVirtual VSA Ready Nodes are configured with two dedicated 1GbE ports for this network function. Additional network adapters can be added if more bandwidth to and from virtual machines is required. No IP addresses are required for the software components on the HPE StoreVirtual VSA Ready Node.
2
Management network
This network segment is used for all management traffic of the system including the hypervisor management functions for each host as well as the HPE iLO 4 server management interface. HPE StoreVirtual VSA Ready Nodes are configured with two dedicated 1GbE ports for this network function. The network function is separate for security and isolation purposes. HPE iLO 4: HPE StoreVirtual VSA: If using VMware vSphere: If using Microsoft Hyper-V:
3
Storage
This network is used for iSCSI storage access (hypervisor to HPE StoreVirtual traffic) as well as internal storage pool communication (HPE StoreVirtual to HPE StoreVirtual traffic). HPE StoreVirtual VSA Ready Nodes are configured with two dedicated 1GbE ports (Small), two 10GbE ports (shared with VMware vSphere® vMotion®/Live Migration in active or standby; Medium, Medium Hybrid) or two dedicated 10GbE ports (Large Hybrid). This network function is typically exposing bursts and should therefore be separated onto its own network segments for network performance concerns. HPE StoreVirtual VSA: If using VMware vSphere: If using Microsoft Hyper-V:
4
vMotion/Live Migration
(1) IP address per HPE StoreVirtual VSA Ready Node (1) IP address per HPE StoreVirtual VSA Ready Node (1) IP address per HPE StoreVirtual VSA Ready Node (1) IP address per HPE StoreVirtual VSA Ready Node + (1) IP address per Windows Failover Cluster
(1) IP address per HPE StoreVirtual VSA Ready Node + (1) per storage pool (2) IP addresses per HPE StoreVirtual VSA Ready Node (1) IP address per HPE StoreVirtual VSA Ready Node
This network is used for virtual machine migration across vSphere hosts. HPE StoreVirtual VSA Ready Nodes are configured with two dedicated 1GbE ports (Small), two 10GbE ports (shared with storage in active or standby; Medium, Medium Hybrid) or two dedicated 10GbE ports (Large Hybrid). This network function is typically exposing bursts and should therefore be separated onto its own network segments for network performance concerns. If using VMware vSphere: If using Microsoft Hyper-V:
(1) IP addresses per HPE StoreVirtual VSA Ready Node (1) IP addresses per HPE StoreVirtual VSA Ready Node
The recommended network port configuration for these network functions in a typical HPE StoreVirtual VSA Ready Nodes configuration is outlined in Table 4. Failover networks teams are implemented in vSphere on separate virtual switches in vSphere, and VLAN separation is implemented on port group and vmkernel. For Windows, a combination of NIC teaming in Windows and Hyper-V Virtual Networks is used.
Technical white paper
Page 10
Table 4. Recommended port mappings and network configuration in vSphere and Windows VLAN
Role
Ports for Small configuration
Ports for Medium and Medium Hybrid configurations
Ports for Large Hybrid configuration
1
VM network
Onboard port 1, FlexLOM port 1
Onboard port 1, Onboard port 2
Onboard port 1, Onboard port 2
vSwitch0 with one port group for virtual machines
vSwitch0 with one port group for virtual machines
vSwitch0 with one port group for virtual machines
Team 1 mapping to Hyper-V Virtual Network
Team 1 mapping to Hyper-V Virtual Network
Team 1 mapping to Hyper-V Virtual Network
Onboard port 2, FlexLOM port 2
Onboard port 3, Onboard port 4
Onboard port 3, Onboard port 4
vSwitch1 with one vmkernel for management, port group for virtual machines optional
vSwitch1 with one vmkernel for management, port group for virtual machines optional
vSwitch1 with one vmkernel for management, port group for virtual machines optional
Team 2, mapping to Hyper-V Virtual Network optional
Team 2, mapping to Hyper-V Virtual Network optional
Team 2, mapping to Hyper-V Virtual Network optional
2
3
Management network
Storage
Onboard port 3, FlexLOM port 3 vSwitch2 with two vmkernel for iSCSI, port group for virtual machines (HPE StoreVirtual VSA will connect to this network segment) Team 3, mapping to Hyper-V Virtual Network (HPE StoreVirtual VSA will connect to this network segment)
4
vMotion/ Live Migration
Onboard port 4, FlexLOM port 4 vSwitch3 with one vmkernel for vMotion and Fault Tolerance Team 4, enabled for Live Migration
FlexLOM port 1, FlexLOM port 2 The ports are used for storage and virtual machine migration purposes vSwitch2 with two vmkernel for iSCSI, port group for virtual machines (HPE StoreVirtual VSA will connect to this network segment), one vmkernel for vMotion Team 3, mapping to Hyper-V Virtual Network, shared with host (HPE StoreVirtual VSA will connect to this network segment), enabled for Live Migration
FlexLOM port 1, NIC port 1, vSwitch2 with two vmkernel for iSCSI, port group for virtual machines (HPE StoreVirtual VSA will connect to this network segment) Team 3, mapping to Hyper-V Virtual Network (HPE StoreVirtual VSA will connect to this network segment) FlexLOM port 2, NIC port 2 vSwitch3 with one vmkernel for vMotion and Fault Tolerance Team 4, enabled for Live Migration
For implementing this networking configuration, see the HPE StoreVirtual VSA design and configuration guide white paper as well as the storage and networking sections in the VMware vSphere or Microsoft Windows Server product documentation.
Storage configuration All HPE StoreVirtual VSA Ready Nodes use a storage configuration to maximize the performance characteristics of the system and the ease of use of the fully configured system. As shown in Figure 8 , the storage configuration is using an HPE S mart Array feature to configure multiple logical drives on a single array of disks. This ensures reliability for the hypervisor boot disk without giving up too many storage resources and allows for exclusive access to the remaining storage capacity by the HPE StoreVirtual VSA.
Figure 8. Graphical representation of the HPE Smart Array configuration.
Figure 9. Storage mapping on HPE StoreVirtual VSA to logical drives.
Technical white paper
Page 11
Logical drives appear to the system as regular disk resources. Except for the hypervisor boot disk, all logical drives are mapped to HPE StoreVirtual VSA as illustrated in Figure 9. In vSphere environments, the logical drives are presented to the HPE StoreVirtual VSA as raw device mappings (RDMs); the additional step required is outlined in the following installation section. Hyper-V supports the use of physical disks on HPE StoreVirtual VSA without any additional steps.
Important Not all arrays and logical drives are created on the same HPE Smart Array controller. On the Medium Hybrid and Large Hybrid configurations, the onboard HPE Smart Array P440ar Controller is used for the SSDs and the additional HPE Smart Array P840 used for the HDDs. See the individual configuration for details (Page 12 for Medium and Page 13 for Medium Hybrid).
HPE StoreVirtual VSA Ready Node—Small The first HPE StoreVirtual VSA Ready Node configuration is targeted toward a typical small office environment with requirements for highly available storage for about 20 virtual machines. While this reference configuration can scale like any other HPE S toreVirtual VSA configuration through scale-out, the recommended HPE StoreVirtual VSA 4 TB license restricts scalability to a s torage pool of three systems. Table 5. HPE StoreVirtual VSA Ready Node—Small 4 TB Quantity
SKU
SKU description
1
TC484AAE
HPE SV VSA 2014 4 TB 3pk 3yr E-LTU
1
755258-B21
HPE DL360 Gen9 8SFF CTO Server
1
818172-L21
HPE DL360 Gen9 E5-2620v4 FIO Kit
1
818172-B21
HPE DL360 Gen9 E5-2620v4 Kit
8
805349-B21
HPE 16 GB 1Rx4 PC4-2400T-R Kit
5
781518-B21
HPE 1.2 TB 12 G SAS 10K 2.5in SC ENT HDD
1
749974-B21
HPE Smart Array P440ar/2G FIO Controller
1
629135-B22
HPE 1 Gb Ethernet 4P 331FLR Adptr
1
734807-B21
HPE 1U SFF Easy Install Rail Kit
2
720478-B21
HPE 500 W FS Plat Ht Plg Pwr Supply Kit
1
P8B31A
HPE OV w/o iLO 3yr 24x7 FIO Phys 1 LTU
1
BD505A
HPE iLO Adv incl 3yr TSU 1-Svr Lic
Important 4 TB licenses come only in license packs of three HPE StoreVirtual VSA licenses. HPE StoreVirtual VSA 4 TB licenses only allow for scaling up to three systems in a storage pool a nd do not support systems with mixed storage configurations with HPE Adaptive Optimization. There are upgrade licenses available to upgrade from 4 TB to the other capacity licenses to enable more capacity in the node and HPE Adaptive Optimization.
The storage configuration is outlined in Table 6. To verify the drive cabling and the other components installed in the server, please see the HPE ProLiant DL360 Gen9 User Guide.
Technical white paper
Page 12
Table 6. Storage configuration for HPE StoreVirtual VSA Ready Node—Small Array #
Controller
Logical drive
RAID type
Size
Purpose
Logical drive 1
RAID 5
1
All drives on HPE Smart Array P440ar
Start with the other logical disk on this array, then assign all remaining capacity to this logical drive
Operating system boot disk (vSphere/Windows)
Logical drive 2
RAID 5
4 TB
HPE StoreVirtual VSA Tier 1 (capacity)
HPE StoreVirtual VSA Ready Node—Medium The Medium configuration is targeted toward larger remote offices with requirements for highly available storage for about 40 virtual machines and larger data sets in remote offices that cannot be centralized easily, such as manufacturing sites and mineral exploration operations. Table 7. HPE StoreVirtual VSA Ready Node—Medium 10 TB Quantity
SKU
SKU description
1
Q0J77AE
HPE StoreVirtual VSA 2014 Software 10 TB E-LTU
1
719064-B21
HPE DL380 Gen9 8SFF CTO Server
1
817937-L21
HPE DL380 Gen9 E5-2640v4 FIO Kit
1
817937-B21
HPE DL380 Gen9 E5-2640v4 Kit
16
805349-B21
HPE 16 GB 1Rx4 PC4-2400T-R Kit
1
768857-B21
HPE DL380 Gen9 8SFF Cage Bay2/Bkpln Kit
12
781518-B21
HPE 1.2 TB 12G SAS 10K 2.5in SC ENT HDD
1
665243-B21
HPE Ethernet 10 Gb 2P 560FLR-SFP+ Adptr
1
783009-B21
HPE DL380 Gen9 8SFF SAS Cable Kit
1
733660-B21
HPE 2U SFF Easy Install Rail Kit
1
843199-B21
HPE Smart Array P840ar/2G Controller
2
720478-B21
HPE 500 W FS Plat Ht Plg Pwr Supply Kit
1
P8B31A
HPE OV w/o iLO 3yr 24x7 FIO Phys 1 LTU
1
BD505A
HPE iLO Adv incl 3yr TSU 1-Svr Lic
The storage configuration is outlined in Table 8. To verify the drive cabling and the other components installed in the server, please see the HPE ProLiant DL380 Gen9 User Guide. Table 8. Storage configuration for HPE StoreVirtual VSA Ready Node—Medium Array #
Controller
Logical drive
RAID type
Size
Purpose
Logical drive 1
RAID 6
1
All HDDs on HPE Smart Array P840ar
Start with the other logical disk on this array, then assign all remaining capacity to this logical drive
Operating system boot disk (vSphere/Windows)
Logical drive 2
RAID 6
10 TB
HPE StoreVirtual VSA Tier 1 (capacity)
Technical white paper
Page 13
HPE StoreVirtual VSA Ready Node—Medium Hybrid Medium Hybrid is a high-performance version of the Medium configuration with 10% of flash storage for increased performance. The design center for this configuration used typical server virtualization in medium-sized businesses with storage requirements for about 60 virtual machines. The configuration is designed to use scale-up capabilities of the HPE StoreVirtual VSA. With the use of two HPE Smart Arra y storage controllers, another set of 10 drives (seven HDDs and three HDDs) can easily be added to the system. When extending the capacity of the HPE StoreVirtual VSA, a license upgrade (HPE S toreVirtual VSA Upgrade 10 TB to 50 TB E-LTU) will be required to virtualize the storage capacity of the additional drives. Table 9. HPE StoreVirtual VSA Ready Node—Medium Hybrid 7.5 TB Quantity
SKU
SKU description
1
Q0J77AE
HPE StoreVirtual VSA 2014 Software 10 TB E-LTU
1
767032-B21
HPE DL380 Gen9 24SFF CTO Server
1
817943-L21
HPE DL380 Gen9 E5-2650v4 FIO Kit
1
817943-B21
HPE DL380 Gen9 E5-2650v4 Kit
8
805351-B21
HPE 32 GB 2Rx4 PC4-2400T-R Kit
3
779168-B21
HPE 400 GB 12G SAS ME 2.5in EM SC H2 SSD
7
781518-B21
HPE 1.2 TB 12G SAS 10K 2.5in SC ENT HDD
1
665243-B21
HPE Ethernet 10 Gb 2P 560FLR-SFP+ Adptr
1
749974-B21
HPE Smart Array P440ar/2G FIO Controller
1
726897-B21
HPE Smart Array P840/4G Controller
2
783009-B21
HPE DL380 Gen9 8SFF SAS Cable Kit
1
786092-B21
HPE DL380 Gen9 8SFF H240 Cable Kit
1
733660-B21
HPE 2U SFF Easy Install Rail Kit
2
720478-B21
HPE 500 W FS Plat Ht Plg Pwr Supply Kit
1
BD505A
HPE iLO Adv incl 3yr TSU 1-Svr Lic
The storage configuration is outlined in Table 10. To verify the drive cabling and the other components installed in the server, please see the HPE ProLiant DL380 Gen9 User Guide. Table 10. Storage configuration for HPE StoreVirtual VSA Ready Node—Medium Hybrid Array #
Storage type
1
All HDDs on HPE Smart Array P840
3
All SSDs on HPE Smart Array P440ar
Logical drive
RAID type
Size
Purpose
Logical drive 1
RAID 5
200 GB
Operating system boot disk (vSphere/Windows)
Logical drive 2
RAID 5
Start with the other logical disk on this array, then assign all remaining capacity to this logical drive
HPE StoreVirtual VSA Tier 1 (capacity)
Logical drive 3
RAID 5
All available capacity
HPE StoreVirtual VSA Tier 0 (performance)
Technical white paper
Page 14
HPE StoreVirtual VSA Ready Node—Large Hybrid The largest HPE StoreVirtual VSA Ready Nodes is the Lar ge Hybrid configuration maxing out the drive configuration of the underlying server platform with 24 drives and a ratio using 18% of flash storage. The processor and memory configuration allows for high-density virtualization projects with about 80 virtual machines per node. Table 11. HPE StoreVirtual VSA Ready Node—Large Hybrid 18.5 TB Quantity
SKU
SKU description
1
Q0J78AE
HPE StoreVirtual VSA 2014 Software 50 TB E-LTU
1
767032-B21
HPE DL380 Gen9 24SFF CTO Server
1
817951-L21
HPE DL380 Gen9 E5-2680v4 FIO Kit
1
817951-B21
HPE DL380 Gen9 E5-2680v4 Kit
16
805351-B21
HPE 32 GB 2Rx4 PC4-2400T-R Kit
6
779172-B21
HPE 800 GB 12G SAS ME 2.5in EM SC H2 SSD
18
781518-B21
HPE 1.2 TB 12G SAS 10K 2.5in SC ENT HDD
1
665243-B21
HPE Ethernet 10 Gb 2P 560FLR-SFP+ Adptr
1
749974-B21
HPE Smart Array P440ar/2G FIO Controller
1
726897-B21
HPE Smart Array P840/4G Controller
2
783009-B21
HPE DL380 Gen9 8SFF SAS Cable Kit
1
786092-B21
HPE DL380 Gen9 8SFF H240 Cable Kit
1
665249-B21
HPE Ethernet 10 Gb 2P 560SFP+ Adptr
1
733660-B21
HPE 2U SFF Easy Install Rail Kit
2
720478-B21
HPE 500 W FS Plat Ht Plg Pwr Supply Kit
1
P8B31A
HPE OV w/o iLO 3yr 24x7 FIO Phys 1 LTU
1
BD505A
HPE iLO Adv incl 3yr TSU 1-Svr Lic
The storage configuration is outlined in Table 12. To verify the drive cabling and the o ther components installed in the server, please see the HPE ProLiant DL380 Gen9 User Guide. Table 12. Storage configuration for StoreVirtual VSA Ready Node—Large Hybrid Array #
Storage type
1
HDD 1–9 on HPE Smart Array P840
Logical drive
RAID type
Size
Purpose
Logical drive 1
RAID 6
200 GB
Operating system boot disk (vSphere/Windows)
Logical drive 2
RAID 6
Start with the other logical disk on this array, then assign all remaining capacity to this logical drive
HPE StoreVirtual VSA Tier 1 (capacity)
2
HDD 10–18 on HPE Smart Array P840
Logical drive 3
RAID 6
All available capacity
HPE StoreVirtual VSA Tier 1 (capacity)
3
All SSDs on HPE Smart Array P440ar
Logical drive 4
RAID 5
All available capacity
HPE StoreVirtual VSA Tier 0 (performance)
Technical white paper
Page 15
Summary As customers are looking to simplify virtualization solutions by using converged compute and storage building blocks, HPE StoreVirtual VSA presents a proven option to deliver data s ervices that seamlessly integrate into virtualized environments. With the heritage of an enterprise storage system, converged solutions based on the HPE StoreVirtual technology work with existing information lifecycle products and other arrays in the HPE storage array fa mily. By selecting HPE StoreVirtual VSA Ready Nodes, customers and par tners can significantly simplify the platform decisions for virtualization solutions and maintain full control over their storage systems.
Appendix A I/O Performance Testing Methodology with Vdbench Introduction The benchmarking of virtualized systems has always proved challenging for a variety of reasons stemming from understanding the network topology and capability, to workload modelling, and scaling for future growth. Commonly, the desire is simply to implement a solution to meet an immediate need, as quickly as possibl e. Shortly thereafter however, the realization sets in that the solution will need to scale in order to continue receiving benefit from the investment already made. There are no shortcuts to designing and implementing an effective solution. Due diligence in planning, p roof of concept/testing, adaptability, and best practice implementation are critical to delivering solutions that satisfy needs. To be successful requires a base level of knowledge on the solution being considered. This knowledge can then be translated into an effective design for a specific use and tailored for the environment. Testing the design provides the necessary knowledge to adapt and grow solutions both for the initial implementation and future growth scaling. This section provides a method to determine the specific level of k nowledge required in order to effectively plan, test, a nd deploy a virtualized solution on either Microsoft Hyper-V or VMware ESXi hypervisors with HPE StoreVirtual VSA that can be replicated and measured using the HPE StoreVirtual VSA Ready Nodes built on HPE ProLiant s erver platforms. It is also possible to replicate these steps on more general x86 platforms in order to baseline performance expectations using the same software stack b ut your performance results will vary based on the components in the platform chosen. We will detail our approach using Vdbench, a freely available testing tool, to benchmark the two medium-sized HPE S toreVirtual VSA Ready Node reference configurations described previously. A complete guide to implementing the test bed and running tests is provided with the intent that partners and customers can replicate HPE’s experience for themselves in their own environments with a reasonable expectation of achieving the same performance results with the reference configurations. Benchmarked reference configurations provide a measured way to build solutions from a modular perspective. A framework for providing solutions that meet costs effectively, and scale predictably. Methodology HPE’s approach to using Vdbench revolves specifically around the reference configurations for the HPE StoreVirtual VSA Ready Nodes. It is important to know in advance that every single component within the reference configurations has the potential to impact performance. Virtual machine memory management, application protocols (i.e., TCP/UDP), disk configuration and mapping, network infrastructure, and compute resources all either add or s ubtract from the ability of any solution to perform to expectations. A s hortfall indicated in a specific area d oes not invalidate the approach but does point out an ar ea that needs to be addressed in order to achieve the desired results. It is for this reason that benchmarking is so critical to successful deployments. If this test process is run on any other hardware/software combination, results will vary; however, the process will still be just as valid for helping to establish that critical base level of knowledge.
Workload Modeling Considerations I/O Queue Depth I/O performance can be increased via multiple simultaneous host requests, as well as, asynchronous applications. Taking advantage of this is of critical importance for the usage of RAID array subsystems li ke the StoreVirtual VSA where a virtual disk is supported by multiple physical disks. The submission of multiple, simultaneous I/Os causes the array to access multiple physical disks simultaneously upping the throughput capability of the SAN.
Technical white paper
Page 16
Transfer Size A variety of block sizes should be used during testing to represent not only the specific size you intend to use as a primary transfer size but also to gauge system performance when a more generalized workload is processed. Applications and operating systems are generally the determining factor in knowing block size and documentation for those items should be r eferenced to understand the solution requirements. Hard drives have their own requirements for sector size, commonly 512 bytes unless using the Advanced Format which could be 4096 bytes as an example and this is specified in their respective product specifications. All I/O operations take a given block of data and either read it from, or write it to some number of disk sectors. There is a direct correlation between block size and I/O throughput or Input/Outputs per Second (IOPS). Changing one impacts the other directly. Example: A single 4 KB block size must access 8 different 512-byte disk sectors; an 8 KB block size would access 16 512-byte disk sectors and so on. Each additional disk sector requires seek time, cache time, etc. Logic may suggest that the smaller block size is more efficient and if the only concern was for the throughput of data at the iSCSI to disk level this would be true but applications insert their own requirements that influence end to end p erformance and efficiency. Read/Write Ratio To completely test a useful range of read/write characteristics the plan incorporates 8k, 16k, 32k, and 256k block sizes in 0%, 60%, 80%, and 100% read configurations. Other block sizes and percentages can be tested optionally if desired but are not part of the provided test plan. The 256k block size equates to a sequential read which i s detailed below. Sequential vs. Random Random reads are the normal mode process for using StoreVirtual to virtualize data for clients. Random is based on some number of clients all sending data packets at unknown intervals to the StoreVirtual cluster. In normal operation the use of the Multipathing Extension Module (MEM) helps to ensure the locality of these packets but does not explicitly reduce the potential for congestion across busy NIC ports or virtual switches without further workload planning for the individual clients or workgroups. Random op erations are more affected by seek time and the rotational latency of the disk (or array) being written to but balance this performance hit with being able to accommodate large numbers of diverse blocks simultaneously. Sequential reads are ordered sets of packets arriving linearly and are thus less dependent on the seek time and rotational latency of the disk (or array). These reads and writes tend to be common with applications for backup and archiving which is why they are included here. It is always good practice to plan for the backup and archiving of your data. Test Environment Configuration The management server for your HPE StoreVirtual VSA Ready Node cluster will need the following additional software components installed with the appropriate permissions and services. 1. Java (jre-8u121-windows-x64) 2. Secure Shell (or the built-in Vdbench version included in the next step) 3. Vdbench (v5.04.06) installed to the root of the administrator’s user space for convenience (other locations are possible but complicate
command line and scripting paths). Each physical server node of the 3-node cluster will run the chosen hypervisor (Hyper-V or ESXi), and have two guest VMs and (4) datastores: 1. Two Windows 2012 R2 datacenter server VMs. Each VM will be created from a 40 GB datastore via the OneView storage management
plug-in. 2. An additional datastore will be provid ed for each VM as the storage repository to be tested and the size varies depending on the cluster
configuration chosen.
Technical white paper
Page 17
Medium-Hybrid Configuration Table 13: Medium-Hybrid Storage Configuration Storage Configuration
Details
HPE StoreVirtual OS version
v12.6
HPE StoreVirtual CMC version
v12.6
Jumbo frames
No
iSCSI offload
No
TOE offload
No
Disk type, Model; firmware version
(3) SSD disks (per node), Model—MO0400JEFPA; Firmware version HPD2 (7) HDD disks (per node), Model—EG1200JEHMC; Firmware version HPD3
Raw capacity per disk/logical array
400 GB/Tier-0, LD-01 RAID 5, 745 GiB 1200 GB/Tier-1, LD -01 RAID 5, 199 GiB 1200 GB/Tier-1, LD -02 RAID 5, 6506 GiB
Controller model, firmware version
Smart Array P480 4.52 Smart Array P440ar 4.52
Flash-backed Write cache (P480)
4194304 KB
Encryption
Not Enabled
Multipath Extension Module (MEM)
MetadataVUM-hpe-lh-mem_6.0.0-12.6.0.14-offline_bundle
Table 14: Medium-Hybrid Host Configuration Host Configuration
Details
hypervisor version
VMware-ESXi-6.0.0-Update2-3620759-HPE-600.9.6.0.49-Oct2016
iLO 4 management, firmware version
Yes 2.50, Sep 23, 2016
Jumbo frames
No
iSCSI offload
No
TOE offload
No
TPM Status
Not Present
SD-Card status
Not Present
ProLiant System ROM, Date
P89 v2.30 Sep 13, 2016
Number of clients
2 per node—Windows 2012 R2
Number of LUNs per client
2 per client
Table 15: Medium-Hybrid Network Switch Configuration Network Switch Configuration
Details
Compute and storage 10GbE switch model, Firmware version
HPE 5900-CP-48XG-4QSFP+
iLO 1GbE switch m odel, firmware version
Technical white paper
Page 18
Medium Configuration Table 16: Medium Storage Configuration Storage Configuration
Details
HPE StoreVirtual OS version
v12.6
HPE StoreVirtual CMC version
v12.6
Jumbo frames
No
iSCSI offload
No
TOE offload
No
Disk type, Model; firmware version
(12) HDD disks (per node), Model—EG1200FDJYT; Firmware version HPD4
Raw capacity per disk/logical array
1200 GB, LD -01 RAID 5, 199 GiB 1200 GB, LD -02 RAID 5, 6506 GiB
Controller model, firmware version
Smart Array P840ar 4.52
Flash-backed Write cache (P840)
2097152 KB
Encryption
Not Enabled
Multipath Extension Module (MEM)
MetadataVUM-hpe-lh-mem_6.0.0-12.6.0.14-offline_bundle
Table 17: Medium Host Configuration Storage Configuration
Details
Hypervisor version
VMware-ESXi-6.0.0-Update2-3620759-HPE-600.9.6.0.49-Oct2016
iLO 4 management, firmware version
Yes 2.50, Sep 23, 2016
Jumbo frames
No
iSCSI offload
No
TOE offload
No
TPM Status
Not Present
SD-Card status
Not Present
ProLiant System ROM, Date
P89 v2.30 Sep 13, 2016
Number of clients
2 per node—Windows 2012 R2
Number of LUNs per client
2 per client—1. OS, 2. Storage
Table 18: Medium Network Switch Configuration Storage Configuration
Details
Compute and storage 10GbE switch model, Firmware version
HPE 5900-CP-48XG-4QSFP+ 7.1.045, Release 2311P01
iLO 1GbE switch m odel, firmware version
HPE OfficeConnect 1950
Technical white paper
Page 19
Vdbench Test Plan and Definitions Create a “parmfile” The parameter file is made up of 3 sections. Storage Definition—Identifies each physical or logical volume or file system file to be used. Workload Definition—Defines what kind of workload will be executed using the storage definitions listed above. Run Definition—Defines what I/O rate to be generated, and how long the workload will run. Notes and Custom Parmfile Definitions: These can be used to run consecutive runs across all three nodes in the cluster with a consolidated output on the management server. hd=default,vdbench=/home/usr/vdbench,user=administrator hd=one,system=system1 hd=two,system=system2 hd=three,system=system3 sd=sd1, host=*,lun=XXXXX wd=wd1, sd=sd*, xfersize=8k, rdpct=100 wd=wd2, sd=sd*, xfersize=8k, rdpct=80 wd=wd3, sd=sd*, xfersize=8k, rdpct=60 wd=wd4, sd=sd*, xfersize=8k, rdpct=0 wd=wd5, sd=sd*, xfersize=32k, rdpct=100 wd=wd6, sd=sd*, xfersize=32k, rdpct=80 wd=wd7, sd=sd*, xfersize=32k, rdpct=60 wd=wd8, sd=sd*, xfersize=32k, rdpct=0 wd=wd9, sd=sd*, xfersize=256k, rdpct=100 wd=wd10, sd=sd*, xfersize=256k, rdpct=80 wd=wd11, sd=sd*, xfersize=256k, rdpct=60 wd=wd12, sd=sd*, xfersize=256k, rdpct=0 rd=run1, wd=wd1,iorate=100, elapsed=XXX Leveraging Performance Results Performance results can be viewed in one of two locations depending on how you approached the testing process but will always be in the “output” directory beneath the installation location for Vdbench itself. For example “C:\Vdbench\output\”. Please note that Vdbench will not automatically create a new output folder with successive tests so it is important to remember to change your original output folder name to something else prior to beginning any new test runs. If you are running tests for multiple nodes from your management server then the output files and data will be located on that system. If however, you have chosen to run the test locally on each node there would be a corresponding “Vdbench/output” folder on that system where the output files and data are located.
Technical white paper
Page 20
The easiest way to engage with the resulting data is to op en the C:/Vdbench/output folder and select the summary.html file.This opens the page in a web browser that conveniently and interactively links to al l of the other collected data that pertains to this particular test run and also indicates if the test plan was completed successfully at the very bottom. The page provides a higher level summary of the test just performed to include: I/O, MB/sec, bytes, read percentage, response time (latency), read response (latency), write response (latency), maximum response time (latency), response time standard deviation, queue depth, and cpu utilization numbers in the bottom half of the page.
Additional depth can be investigated by selecting a link from the top half of the page. Selecting the “logfile” link will show a detailed and time stamped log which is useful in the event tests are not completing as expected for troubleshooting. The “status” link can be selected at any time after the test has been initiated to see where in the process the test is currently. The “parmfile” link will show you the specific parameters used to perform the test and is dependent on the definitions provided by the user. The link to “host reports” would show each of the hosts included in the test individually if more than one was included i n the parmfile used to run the test. The “skew” report is of benefit only after having run multiple workloads with different characteristics (i.e. r/w %, blk sizes, etc.) on multiple hosts to compare results across the different workloads and hosts for consistency. The “SD reports” will show for each storage definition (“sd”) defined in the parmfile and can be accessed individually via these links here. Likewise, the different workloads defined (“wd”) in the parmfile will be listed and can be accessed in the workload report sections. Lastly, the run definitions (“rd”) are also listed and can be selected. Assuming the tests ran completely and successfully the most important items to focus on are the ‘‘summary.html’’; the ‘‘totals’’; ‘‘histogram’’; and the ‘‘skew’’ reports; with additional information being leveraged from the individual host reports only if the skew report indicates significant differences on any of the hosts.
Technical white paper
Page 21
Appendix B Baseline Performance Results Keeping in mind that performance will vary according to settings and network environment the following graphs demonstrate the HPE StoreVirtual Ready Node Medium configuration results for a variety of block sizes and Read/Write (R/W) percentages. We focus on the 8k and 32k block sizes for random reads and writes and then finish with the 256k sequential results that should be relevant when considering backup performance expectations. The blue line represents our all spinning disk (HDD) configuration results and the yellow line represents our tier-0 SSD tier p erformance on the Medium Hybrid configuration. All results from one of two 3-node DL38 0 Gen9 server clusters running VMware ESXi 6.0U2 and HPE StoreVirtual VSA 12.6. IO/sec vs. RT(ms) random; Blocksize=8192; %R/W-Ratio=100
45 40 35 30
) s m25 ( T R 20 15 10 5 0 0
10000
20000
3000 0
40000
5000 0
6000 0
7000 0
8000 0
90000
IO/sec VSA_ReadyNode_NR10_R6_36HDD_1.2 TB_iSCSI10GbE_v12.6_RR_8K_T1x.xlsx:random; Blocksize=8192; %R/W-Ratio=100; DeviceCount:6
VSA_ReadyNo de_Hybrid-Cluster__R R_8K—Copyx.xlsx: random; Blocksize=8192; %R/W-Ratio=100; DeviceCount:6
100000
Technical white paper
Page 22
IO/sec vs. RT(ms) random; Blocksize=8192; %R/W-Ratio=80
45 40 35 30
) s m 25 ( T R 20 15 10 5 0 0
10000
20000
30000
40000
50000
60000
70000
IO/sec VSA_ReadyNode_NR10_R6_36HDD_1.2 TB_iSCSI10GbE_v12.6_RMIX80_8K_T1x.xlsx:random; Blocksize=8192; %R/W-Ratio=80; DeviceCount:6
VSA_ReadyNode_Hybrid-Cluster__RMIX80_8K2—Copyx.xlsx:random; Blocksize=8192; %R/W-Ratio=80; DeviceCount:6
IO/sec vs. RT(ms) random; Blocksize=32768; %R/W-Ratio=100
45 40 35 30 ) 25 s m ( T20 R
15 10 5 0 0
10000
20000
30000
40000
50000
60000
70000
IO/sec VSA_ReadyNode_NR10_R6_36HDD_1.2 TB_iSCSI10GbE_v12.6_RR_32K_T1x.xlsx:random; Blocksize=32768; %R/W-Ratio=100; DeviceCount:6 VSA_ReadyNode_Hybrid-Cluster__RR_32K—Copyx.xlsx:random; Blocksize=32768; %R/W-Ratio=100; DeviceCount:6
80000
Technical white paper
Page 23
IO/sec vs. RT(ms) random; Blocksize=32768; %R/W-Ratio=60
50 45 40 35 30 ) s m25 ( T R
20 15 10 5 0 0
5000
10000
15000
20000
25000
30000
35000
40000
45000
IO/sec VSA_ReadyNod e_NR10_R6_36HDD_ 1.2 TB_iSCSI10G bE_v12.6_RMIX6 0_32K_T1x.xlsx:r andom; Blocksize=32 768; %R/W-R atio=60; DeviceCount:6 VSA_ReadyNode_Hybrid-Cluster__RMIX60_32K2—Copyx.xlsx:random; Blocksize=32768; %R/W-Ratio=60; DeviceCount:6
IO/sec vs. RT(ms) random; Blocksize=32768; %R/W-Ratio=80
45 40 35 30 ) 25 s m ( T20 R
15 10 5 0 0
10000
20000
30000
40000
50000
IO/sec VSA_ReadyNode_NR10_R6_36HDD_1.2 TB_iSCSI10GbE_v12.6_RMIX80_32K_T1x.xlsx:random; Blocksize=32768; %R/W-Ratio=80; DeviceCount:6 VSA_ReadyNode_Hybrid-Cluster__RMIX80_32K2—Copyx.xlsx:rand om; Blocksize=32768; %R/W-Ratio=80; DeviceCount:6
60000
Technical white paper
Page 24
IO/sec vs. RT(ms) Sequential; Blocksize=262144; %R/W-Ratio=100
50 45 40 35 30 ) s m25 ( T R
20 15 10 5 0 0
2000
4000
6000
8000
10000
12000
14000
IO/sec VSA_ReadyNode_NR10_R6_36HDD_1.2 TB_iSCSI10GbE_v12.6_SR_256K_T1x.xlsx:Sequential(segmented-method); Blocksize=262144; %R/W-Ratio=100; DeviceCount:6
VSA_ReadyNode_Hybrid-Cluster_SR_256K—Copyx.xlsx:Sequential(segmented-method); Blocksize=262144; %R/W-Ratio=100; DeviceCount:6
Technical white paper
Page 25
Reference and additional resources This section is an overview of all the r esources used in this document. See the following documentation for more background or detail on product features and steps described at a very high level earlier in this document. Table 19. List of rel evant documents and product documentation Product
HPE StoreVirtual VSA
HPE ProLiant Server
HPE OneView for vCenter
HPE OneView for System Center
VMware vSphere Windows Server
Document
Link
HPE StoreVirtual VSA QuickSpecs
hpe.com/h20195/V2/GetDocument.aspx?docname=c04111621
HPE StoreVirtual Architecture
hpe.com/h20195/V2/GetDocument.aspx?docname=4AA3-0365ENW
HPE StoreVirtual VSA design and configuration
hpe.com/h20195/V2/GetDocument.aspx?docname=4AA4-8440ENW
HPE StoreVirtual documentation
hpe.com/storage/docs/storevirtual
HPE StoreVirtual VSA downloads
hpe.com/storage/downloads/storevirtual
StoreVirtual Peer Motion
hpe.com/h20195/V2/GetDocument.aspx?docname=4AA4-0304ENW
HPE StoreVirtual Storage network design considerations and best practices
hpe.com/h20195/V2/GetDocument.aspx?docname=4AA2-5615ENW
HPE ProLiant DL360 Gen9 User Guide
hpe.com/support/dl360gen9_ug_en
HPE ProLiant DL380 Gen9 User Guide
hpe.com/support/dl380gen9_ug_en
HPE ProLiant Recipe for VMware vSphere
vibsdepot.hpe.com/hpq/recipes/HPE-VMware-Recipe.pdf
Service Pack for ProLiant Quick Start Gui de
hpe.com/support/SPP_UG_en
Service Pack for ProLiant download
hpe.com/servers/spp/download
Integrated Lights-On documentation
hpe.com/support/ilo4_ug_en
Intelligent Provisioning documentation
hpe.com/support/IPGen9_UG_en
HPE Smart Storage Administrator User Guide
hpe.com/info/smartstorage/docs
HPE OneView for vCenter information
hpe.com/us/en/product-catalog/detail/pip.4152978.html
HPE OneView for vCenter download
h20392.www2.hpe.com/portal/swdepot/displayProductInfo.do?productNumber=HPVPR
HPE OneView for vCenter documentation
hpe.com/info/ovvcenter/docs
HPE OneView for System Center information
hpe.com/us/en/product-catalog/detail/pip.5390822.html
HPE OneView for System Center download
h20392.www2.hpe.com/portal/swdepot/displayProductInfo.do?productNumber=System_Center
HPE OneView for System Center documentation
hpe.com/info/ovsc/docs
HPE custom VMware® ESXi™ ISO images
hpe.com/info/esxidownload
VMware vSphere 6.0 documentation
pubs.vmware.com/vsphere-60/index.jsp
Windows Server 2012 R2 documentation
technet.microsoft.com/en-us/library/hh801901
Technical white paper
Make the right purchase decision. Click here to chat with our presales specialists.
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. Intel Xeon is a trademark of Intel Corporation in the U.S. and other countries. Microsoft, Windows, and Windows Server are either registered trademarks or trademarks of Microsoft Corporation in the United States and/or other countries. Linux is the registered trademark of Linus Torvalds in the U.S. and other countries. VMware vSphere, VMware vCenter, and VMware ESXi are registered trademarks or trademarks of VMware, Inc. in the United States and/or other jurisdictions. Java is a registered trademark of Oracle and/or its affiliates. All other third-party trademark(s) is/are property of their respective owner(s) a00000521ENW, August 2017, Rev.1