Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users A Dell Compellent VDI Reference Architecture Dell Compellent Technical Solutions December 2013

Revisions Date Jan 2014 Initial publication Damon Zaylskie, Chhandomay Mandal THIS WHITE PAPER IS FOR INFORMATIONAL PURPOSES ONLY, AND MAY CONTAIN TYPOGRAPHICAL ERRORS AND TECHNICAL INACCURACIES. THE CONTENT IS PROVIDED AS IS, WITHOUT EXPRESS OR IMPLIED WARRANTIES OF ANY KIND. 2013 Dell Inc. All rights reserved. Reproduction of this material in any manner whatsoever without the express written permission of Dell Inc. is strictly forbidden. For more information, contact Dell. PRODUCT WARRANTIES APPLICABLE TO THE DELL PRODUCTS DESCRIBED IN THIS DOCUMENT MAY BE FOUND AT: http://www.dell.com/learn/us/en/19/terms-of-sale-commercial-and-public-sector Performance of network reference architectures discussed in this document may vary with differing deployment conditions, network loads, and the like. Third party products may be included in reference architectures for the convenience of the reader. Inclusion of such third party products does not necessarily constitute Dell s recommendation of those products. Please consult your Dell representative for additional information. Trademarks used in this text: Dell, the Dell logo, Dell Boomi, Dell Precision,OptiPlex, Latitude, PowerEdge, PowerVault, PowerConnect, OpenManage, EqualLogic, Compellent, KACE, FlexAddress, Force10 and Vostro are trademarks of Dell Inc. Other Dell trademarks may be used in this document. Cisco Nexus, Cisco MDS, Cisco NX- 0S, and other Cisco Catalyst are registered trademarks of Cisco System Inc. EMC VNX, and EMC Unisphere are registered trademarks of EMC Corporation. Intel, Pentium, Xeon, Core and Celeron are registered trademarks of Intel Corporation in the U.S. and other countries. AMD is a registered trademark and AMD Opteron, AMD Phenom and AMD Sempron are trademarks of Advanced Micro Devices, Inc. Microsoft, Windows, Windows Server, Internet Explorer, MS-DOS, Windows Vista and Active Directory are either trademarks or registered trademarks of Microsoft Corporation in the United States and/or other countries. Red Hat and Red Hat Enterprise Linux are registered trademarks of Red Hat, Inc. in the United States and/or other countries. Novell and SUSE are registered trademarks of Novell Inc. in the United States and other countries. Oracle is a registered trademark of Oracle Corporation and/or its affiliates. Citrix, Xen, XenServer and XenMotion are either registered trademarks or trademarks of Citrix Systems, Inc. in the United States and/or other countries. VMware, Virtual SMP, vmotion, vcenter and vsphere are registered trademarks or trademarks of VMware, Inc. in the United States or other countries. IBM is a registered trademark of International Business Machines Corporation. Broadcom and NetXtreme are registered trademarks of Broadcom Corporation. Qlogic is a registered trademark of QLogic Corporation. Other trademarks and trade names may be used in this document to refer to either the entities claiming the marks and/or names or their products and are the property of their respective owners. Dell disclaims proprietary interest in the marks and names of others. 2 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

Table of contents Revisions... 2 Executive summary... 5 1 Introduction... 6 1.1 Objectives... 6 1.2 Audience... 6 2 Solution test environment... 7 2.1 Software... 7 2.1.1 VDI with XenDesktop... 7 2.1.2 XenDesktop components... 8 2.1.3 VMware components... 8 2.2 Hardware... 9 2.2.1 Storage... 9 2.2.2 Hosting environment... 10 2.2.3 Networking... 12 3 Test methodology...14 3.1 Test objectives...14 3.2 Test tools...14 3.2.1 Load generation...14 3.2.2 Monitoring tool... 15 3.3 Test phases... 15 3.3.1 Boot storm... 15 3.3.2 Login storm...16 3.3.3 Steady state...16 3.4 Test criteria...16 3.4.1 Storage throughput and latency...16 3.5 Test configuration... 17 3.6 Test results... 17 3.6.1 Boot storm... 18 3.6.2 Login storm...19 3.6.3 Steady state... 21 4 Conclusion... 23 3 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

A Appendix... 24 A.1 PVS... 24 A.2 XenDesktop... 24 A.3 VMware ESXi Hypervisor... 24 A.4 VMware vcenter... 25 A.5 LoginVSI... 25 B Terminology... 26 4 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

Executive summary Desktop virtualization is an important strategy for organizations seeking to reduce the cost and complexity of managing an expanding variety of client desktops, laptops, and mobile handheld devices. Virtual Desktop Infrastructure (VDI) offers an opportunity to not only reduce the operational expenses for desktop management and provisioning but also to improve user mobility and data security. A VDI deployment can place high performance and capacity demands on the storage platform. For example, consolidating large amounts of inexpensive stand-alone desktop storage into a centralized infrastructure can create tremendous capacity demands on centrally managed shared storage used in VDI deployments. Performance demands are determined by the number of I/O operations per second (IOPS) generated by basic desktop client operations such as system boot, logon and logoff, and by desktop usage operations from different users. Storm events such as morning logons and afternoon logoffs by many users at approximately the same time can cause I/O spikes that place high performance demands on the storage infrastructure. Moreover, IOPS generated per desktop can vary greatly depending on the user type. For example, knowledge workers with several demanding applications can create significantly higher IOPS per desktop than task workers performing routine operations. A cost-effective, appropriately-sized storage platform is critical for a VDI deployment success. The Dell Compellent SC8000 All Flash Array (SC8000 AFA) with its advanced storage architecture and feature set, combined with tight hypervisor integration, provides an automated, self-tuning storage system capable of efficiently scaling to satisfy VDI production-level workloads. This paper demonstrates how 3,000 virtual desktops can be deployed in a Citrix XenDesktop environment leveraging a single SC8000 AFA comprising two controllers and one enclosure of 24 Solid State Drives (SSDs). The supported VDI workload was medium to heavy, with each desktop generating 30 IOPS at steady state. The SC8000 AFA delivered these IOPS requirements 90,000 at steady state with less than 3.5 ms of storage latency, indicating an excellent experience for the end-users. Details are also provided for the storage I/O characteristics under various other VDI workload scenarios like boot and login storms. 5 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

1 Introduction A storage platform like the Dell Compellent SC8000 All Flash Array (SC8000 AFA) makes an excellent storage platform for VDI. When used in conjunction with Citrix XenDesktop and Citrix Provisioning Server (PVS), layered on top of VMware ESXi server, large scale deployments can be simplified and made more cost effective. The goal of this document is to provide a storage architecture reference point for building large scale VDI deployments with medium to heavy workloads from a storage perspective. The simulation software used simulates a real world implementation. The following software was used: Citrix XenDesktop 5.6 Citrix Provisioning Server VMware ESXi VMware vsphere management LoginVSI (for load management) The hardware used to host the environment consisted of the following: Dell PowerEdge M620 blade servers for VDI workload Dell PowerEdge M710HD blade servers for infrastructure hosting Dell PowerConnect 8024-k blade switches Dell Force10 S4810 Top of Rack (ToR) switch Brocade 5100 Fibre Channel (FC) ToR switches Dell Compellent SC8000 storage array The hardware components are fully detailed in Section 2.2. 1.1 Objectives The primary objectives of the tests conducted for this paper were: Determine the performance impact on the storage array during the peak VDI I/O activity such as boot and login storms Show the latency at the full steady-state user load of 30 IOPS per desktop 1.2 Audience This document is intended for anyone who wishes to understand the performance characteristics of an SC8000 AFA in a VDI solution. This could include, but is not limited to storage architects, customers, solutions partners, and enterprise technologists. 6 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

2 Solution test environment The tested environment was designed to maximize the storage load. The hardware external to the storage array was greater than required for the user load to ensure that it was not a limiting factor. In an actual implementation, the hardware required to host and manage the user load may not match the actual hardware used for these tests. The goal was to determine the maximum performance the storage array can deliver in the VDI environment without any other component in the stack being the bottleneck. 2.1 Software Since the goal was to simulate 3,000 knowledge workers running real world applications like Microsoft Office and Adobe Acrobat, the actual management platforms were simulated as well. 2.1.1 VDI with XenDesktop XenDesktop provides a highly scalable platform for VDI. For enterprise customers wishing to deploy large scale virtualized environments, it provides a highly scalable, modular architecture that can be easily adapted to fit almost any environment. XenDesktop offers several options for VDI connectivity. This testing used the Citrix flash-based web interface for ease of testing and dynamic load balancing of connections. This provides a single point of entry for remote connections and automatically balances the load across front-end servers while connections are initiated. 7 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

2.1.2 XenDesktop components The table below shows the components of XenDesktop that were used in the testing. Table 1 XenDesktop components Desktop Delivery Controller XenDesktop Web PVS Virtual Desktop Agent PVS Virtual Disk agent Client Devices Citrix Receiver The management platform for controlling desktops and assigning privileges The component that provides the web portal for remote clients The component that streams the operating system to the virtual machines and includes the dynamic load balancing of boot streams Manages virtual machine registration and provides centralized management. The agent registers with the XenDesktop servers and reports the client state as well as performance and environment data Provides details on each boot device. This agent shows read/write data, errors, and state of the boot device For this testing, Windows 2008 R2 launchers were used to initiate the virtual machine sessions using the Citrix Receiver client and XenDesktop Web Client program that connects to virtual machines managed by XenDesktop which is initiated through XenDesktop Web 2.1.3 VMware components The table below shows the VMware components used during the testing. Table 2 VMware components vsphere Client vcenter Server ESXi server VMware Tools The client for accessing the vcenter servers The server that manages the ESXi hosts and holds a central repository for ESXi server data and configuration The hypervisor host on which the virtual machines run An agent that runs in each virtual machine used by vcenter to manage the client state and collect health data Virtual machines The virtual machines available for remote client access. Each virtual machine is a member of the general pool of available resources vmotion The vmotion feature allows dynamic load balancing of the virtual machines to prevent a situation where one host has a disproportionate virtual machine load based on memory and CPU usage in the hosted virtual machines 8 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

2.2 Hardware 2.2.1 Storage The hardware used to simulate the environment is described in detail in this section. The VDI storage was an SC8000 AFA running version 6.4 firmware. All front end ports were 8 Gb FC and all back end ports were 6 Gb serial attached iscsi (SAS). The array used one enclosure of solid-state drives (SSDs). All drives were 400 GB Single-level Cell (SLC) drives composed of 23 active drives with one hot spare. The infrastructure storage was an SC8000 AFA running version 6.3 firmware. All front end ports were 8 Gb FC and all back and ports were 6 Gb SAS. The array used a combination of SSDs and spinning media. The role of the infrastructure storage is purely to provide storage for domain controllers, launchers, XenDesktop servers, and other miscellaneous server roles. All storage hardware was configured to Dell Compellent best practices with vsphere 5.x. For more information on these Best Practices, see Dell Compellent Best Practices with VMware vsphere 5.x http://www.dell.com/learn/us/en/19/business~solutions~whitepapers~en/documents~compellent-bestpractices-vmware-vsphere-5x.pdf Table 3 Storage hardware components Storage role Type Quantity Controllers VDI workload SC8000 2 System Center Operating System (SCOS) 6.4.1 Enclosures SC220 1 24 bay 2.5 disk drive enclosure Ports FC - 8 Gbps SAS - 6 Gbps 16 4 Front end host connectivity Back end drive connectivity Drives 400 GB SLC drives 24 23 Active with 1 hot spare These are the volumes created on the Storage Center to store virtual machine data for the desktops. Table 4 Virtual Machine Storage Name Type Quantity Size VDI Volume: Write Cache VMFS 60 2 TB Each thin provisioned volume stores 50 virtual machine temp cache VMDKs as write cache 9 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

Table 5 VMWare vswitch configurations Item VLAN Ports NIC Connectivity NIC Teaming mode vswitch Cluster 1 31 4088 Dual 10 Gb NICs, LACP trunked Load Balancing, Route based on IP Hash vswitch Cluster 2 41 4088 Dual 10 Gb NICs, LACP trunked Load Balancing, Route based on IP Hash vswitch Cluster 3 51 4088 Dual 10 Gb NICs, LACP trunked Load Balancing, Route based on IP Hash Infrastructure Chassis 550 1024 Dual 10 Gb NICs, LACP trunked Load Balancing, Route based on IP Hash Each VDI workload chassis is a member of a distinct VLAN, to control the number of devices per VLAN. This allows scaling of the number of VMs per VLAN up to 4,000 per VLAN, based on the vswitch configuration and using S4810 switches, which have a limit of 4,000 Address Resolution Protocol (ARP) entries. 2.2.1.1 PVS device cache layout (write cache for target devices) For each virtual machine stored on the VMFS datastore, the write cache files are as follows. Table 6 PVS local drives Name Type Qty Size Consumed Write cache disk VMDK 1 30 GB ~800-900 MB Stores delta changes from vdisk (thin provisioned) Miscellaneous VM config files MISC ~12 N/A ~1.4 MB Miscellaneous supporting VM config files (vmx, log, nvram, vmxf, hlog, vmsd) 2.2.2 Hosting environment The hosting environment ran VMware ESXi. All hosts were blade chassis based, providing full hardware remote control and power management, as well as simplifying hardware updates, switch management, and standardizing hardware from a replacement perspective. Chassis based servers also simplify cluster management, with a chassis based networking backplane providing high-performance, redundant connectivity. Figure 1 shows the physical networking diagram. 10 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

Figure 1 Physical connectivity 11 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

Table 7 Host and desktop virtual machine configuration Server M620, 320 GB RAM, 2x8 core E5-2680 @ 2.7 Gz Operating system VMware ESXi 5.1 Update 1 FC QLogic QME2572 HBA Network PowerConnect 8024-k, Brocade 8424 Virtual desktop hardware 2vCPUs, 3 GB RAM, VMXNET3, 30 GB cache disk Table 8 Infrastructure virtual machine configuration Operating system vnic Driver version Applications Windows Server 2008 R2 VMXNET3 Active Directory Domain Controllers, DNS, DHCP, vsphere servers, Enterprise Manager server, Liquidware Labs Stratusphere UX, XenDesktop, Microsoft SQL Table 9 PVS configuration Server type XenDesktop Physical, Dell M710HD, 2x6 core processor, 288 GB RAM 5.6 FP1 PVS 6.1 Storage 146 GB local boot volume, 500 GB SAN data volume 2.2.3 Networking Table 10 Network switches ToR switch Force10 S4810 Quantity 1 Firmware 9.2.2 Applications ToR switch, VLAN management, VLAN routing 12 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

Table 11 Infrastructure Blade Chassis network switches Chassis switch Force10 MXL 10 Gb Ethernet Quantity 2 Firmware 9.2.2 Application Chassis based switching for Launcher hosts and PVS Table 12 Workload Blade Chassis network switches Chassis switch Quantity Dell PowerConnect 8024-k 10 Gb Ethernet 6 (2 per blade chassis) Firmware 5.0.1.3 Applications Chassis based switching, server connectivity Table 13 Blade Chassis FC Switches Chassis switch Brocade 5424 Quantity Firmware Applications 6 (2 per blade chassis) 7.1.0.a Chassis based switching, storage connectivity Table 14 ToR FC switches FC switch Brocade 5100 Quantity 2 Firmware Applications 7.1.0a FC fabric, chassis and storage connectivity 13 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

3 Test methodology This section outlines the testing process, criteria, tools, and monitoring methods. The criteria defined here can be used to compare to a customer configuration or desired architecture. 3.1 Test objectives As stated previously, the primary objectives of the tests conducted for this paper were: Determine the performance impact on the storage array during the peak VDI I/O activity such as boot and login storms Show the latency at the full steady state user load of 30 IOPS per desktop The goal of this document is to provide a storage architecture reference point for building large scale VDI deployments with medium to heavy workloads from a storage perspective. 3.2 Test tools In this project, multiple tools were required to control the test. Because there is no single workload tool that can manage the workload and virtual machines from a single interface, a group of tools was chosen to provide the control. XenDesktop was used to throttle the virtual machine boot rate. For the purposes of this test, a variable rate of machines was booted per minute. Since XenDesktop is normally used to boot the virtual machines before they are actually needed using a scheduler, the boot storm phase was purely an IOPS/latency exercise. Since the workload was generated across a large number of machines in a consistent manner, LoginVSI was used to control the login rate for the workstations. By testing varying rates of machines per minute that login, the maximum login rate per minute that the environment can handle was determined. Iometer was used to increase the IOPS per desktop being generated to meet the requirements of the test. LoginVSI only generates approximately 7.5 IOPS/desktop under normal operation, so to supplement the remaining I/O deficit, Iometer was used to generate the additional IOPS. 3.2.1 Load generation The load generation was done with a combination of LoginVSI and Iometer. The following configuration was tested: Table 15 Load generation configuration Workload LoginVSI Iometer Total Power User 7.5 IOPS 22.5 IOPS 30 IOPS 14 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

The characteristics of the LoginVSI Medium workload, designed to run on two vcpus per desktop VM in LoginVSI 4.0, are: Once a session has been started, the workload will repeat (loop) every 48 minutes. The loop is divided in 4 segments; each consecutive LoginVSI user logon will start at different segments. This ensures that all elements in the workload are equally used throughout the test. During each loop the response time is measured every 3-4 minutes. The medium workload opens up to 5 applications simultaneously. The keyboard type rate is 160 ms for each character. Approximately 2 minutes of idle time is included to simulate real-world users. In every loop, each user will open and use: Microsoft Outlook to browse messages. Microsoft Internet Explorer to browse different webpages; a YouTube style video (480p movie trailer) is opened three times in every loop. Microsoft Word to measure response time in one instance, in another instance to review and edit a document. Doro PDF Printer & Acrobat Reader to print the word document and review in PDF. Microsoft Excel to open a very large randomized worksheet. Microsoft PowerPoint to review and edit a presentation. FreeMind, a Java based Mind Mapping application. Iometer generates I/O to a test file, and in this scenario it was a purely random workload of 100% writes. The purpose is to generate a continuous high IOPS load across all machines. This, combined with the LoginVSI workload, generated a workload similar to an actual workstation. 3.2.2 Monitoring tool The performance and latency measuring were done with Dell Compellent Enterprise Manager. This tool provided consolidated performance data across the entire environment from a single console. 3.3 Test phases The VDI environment has multiple types of load, each with a distinct profile. This poses a challenge from a sizing and cost perspective because the need to support high boot loads needs to be balanced against the overall cost of the solution. 3.3.1 Boot storm The boot storm was the phase of the test where all of the virtual machines were turned on in rapid succession. The boot storm was measured from the time the first machine was turned on until the time the last machine was available for login. 15 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

In this test scenario the boot time was limited by the streaming performance of the PVSs. This is due to the fact that the machines were booting across the network from the PVSs. The boot rate was limited by the network connections of the PVSs themselves. 3.3.2 Login storm The login storm phase was designed to test the storage array performance when all 3,000 users were logged into the virtual machines. A high login rate was used to generate substantial I/O load during this login storm phase. 3.3.3 Steady state The steady-state performance was the primary focus of this test. The goal was to achieve the lowest possible latency during the steady-state phase of the test with 3,000 users. The majority of the usage of a VDI solution is providing users with centrally managed workstations, so the most important criteria for this solution is providing the best possible user experience. 3.4 Test criteria There were several test criteria that needed to be monitored during the testing to show the performance of the system under load. 3.4.1 Storage throughput and latency The goal for this test was to maintain under 10 ms of volume latency. This is to highlight the performance characteristics of solid state drives. 16 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

3.5 Test configuration The virtual machines are configured in six Collections within PVS. Each collection contains 500 virtual machines. Table 16 Testing details Item Count PVS 7 Servers that provide the streaming boot image for the virtual machines Device Collections 6 Logical collections of machines for management purposes XenDesktop Servers 2 Desktop Controllers and web access controllers Virtual machines 3,000 The virtual machines being tested VM Datastores 60 The volumes for virtual machine storage and virtual machine disk cache VM Workload hosts VM Infrastructure hosts LoginVSI Launchers 48 ESXi hosts 12 ESXi hosts 150 The machines initiating the remote connections to the virtual machines Table 17 LoginVSI configuration Configuration option Setting Launchers 150 Connection Citrix Receiver web initiated User count 3,000 Workload Run time Medium 4 hours 3.6 Test results The following sections detail the test results in each of the phases of the tests. In each section there is a screenshot showing the actual performance followed by a detailed description of the results. 17 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

3.6.1 Boot storm The boot storm phase of the tests is the least common scenario. The boot storm only happens if the entire system is restarted, such as if it is taken down for maintenance or a power failure or other catastrophic error occurs. Figure 2 Boot storm I/O The maximum I/O achieved was approximately 30,000 IOPS. This is due to the fact that the majority of the I/O was coming from the PVSs, since the PVSs provide all of the boot data only temporarily and swap data is written to disk during this phase. 18 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

Figure 3 Network usage during login storm Figure 3 shows the network load during the boot phase. This load is distributed across seven PVSs. 3.6.2 Login storm The login storm phase is a much more common scenario. This scenario occurs every time the workday starts and is a more important test. The following charts show the performance during the login phase. 19 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

Figure 4 Login storm I/O chart Figure 4 shows the entire login to steady-state transition and highlights the maximum I/O achieved during this test. There is a brief spike in the latency at the end of the login phase where the steady-state transition occurs. The write latency during this time does not exceed 10 ms, and this shows that the users will experience very good performance during their login. 20 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

Figure 5 Maximum I/O during the login phase Figure 5 shows in more detail that the maximum I/O achieved was 93,000 IOPS with a read/write ratio of 5% read and 95% write. 3.6.3 Steady state The steady-state phase is an important phase of the test cycle. The majority of the system time will be spent in steady-state, and the performance during this phase of the test is the most critical because it determines the user experience. Figure 6 shows the performance during the steady-state phase of the test. The two SC8000 controllers with 24 SLC SSDs within one SC220 enclosure consistently delivered 90,000 IOPS for the 3,000 user VDI environment. The average write latency during this test was approximately 3.5 ms. This equates to an excellent user experience during the steady-state phase. In a PVS environment the read latency becomes somewhat irrelevant when the number of read IOPS approaches zero. 21 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

Figure 6 Steady state I/O 22 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

4 Conclusion The paper demonstrates how a 3,000 user XenDesktop VDI environment can be deployed with a single SC8000 AFA comprising two controllers and one enclosure of 24 SLC SSDs. The storage I/O characteristics under various VDI workload scenarios boot storm, login storm and steady state demonstrate the effectiveness of SC8000 AFA as a storage building block for desktop virtualization deployments. The testing of the SC8000 AFA storage platform showed that it can easily support medium to heavy VDI workloads. For example, it delivered 90,000 IOPS with less than 3.5 ms latency at 30 IOPS/desktop for this 3,000 user deployment. With the ability to support a large number of desktops with a small footprint, the SC8000 AFA can help organizations that are limited by the cost-per-desktop hurdle start their VDI deployment plans much sooner. With its superb performance, advanced storage architecture and feature set, combined with tight hypervisor integration, SC8000 AFA provides an excellent platform for VDI infrastructure. 23 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

A Appendix A.1 PVS Table 18 PVS configurations PVS 6.1 Application Server Storage Virtual Machine streaming boot, boot device management, boot versioning 2x6 core processor, 288 GB RAM 146 GB local boot volume, 500 GB SAN data volume A.2 XenDesktop Table 19 XenDesktop server configurations XenDesktop Application Server Storage 5.6 FP1 Virtual machine assignment management, Web interface for remote connections, 2x6 core processor, 288 GB RAM 146 GB local boot volume, 500 GB SAN data volume A.3 VMware ESXi Hypervisor Table 20 VMWare ESXi server configuration ESXi 5.1 Update 1 Application Server Storage End-User Virtual Machine host 2x8 core processor, 320 GB RAM 200 GB SSD local boot volume, SAN based VMFS volumes 24 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

A.4 VMware vcenter Table 21 VMWare vcenter configuration VMware vcenter 5.1 Update 1 Application Hypervisor management, centralized management A.5 LoginVSI Table 22 LoginVSI details LoginVSI 4.0.7 Application Load generation test harness, controls all virtual machine sessions used for load generation 25 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users

B Terminology The following is a list of abbreviations used in this document. IOPS: input output operations per second, disk reads and writes PVS: provisioning server VDI: virtualized desktop infrastructure VMDK: virtual disk stored in a data store VMFS: VMware data store volume 26 Citrix XenDesktop with Provisioning Server for VDI on Dell Compellent SC8000 All Flash Arrays for 3,000 Users