Cisco Integrated Desktop Virtualization Solution

Cisco Integrated Desktop Virtualization Solution VMware View 5.1 with VMware vsphere 5 Using Cisco Unified Computing System, Cisco Nexus Switches, and Nexenta Storage February 2013 2013 Cisco Nexenta. All rights reserved. Page 1

Contents Objectives... 4 Summary of Main Findings... 4 Business Value... 4 Modular Virtual Desktop Infrastructure Overview... 5 Cisco Data Center Infrastructure for Desktop Virtualization... 5 Simplified... 5 Secure... 5 Scalable... 5 Savings and Success... 5 Project Planning and Solution Sizing Sample Questions... 6 The Solution: A Unified, Pretested, and Validated Infrastructure with Cisco, Nexenta, and VMware-Based Reference Architecture... 7 Base Components... 7 Cisco Unified Computing System... 7 Cisco Nexus 5000 Series Switches... 8 VMware vsphere 5... 8 VMware View 5... 8 NexentaVSA for View... 8 Solution Overview and Benefits... 9 Solution Benefits... 10 Architecture Overview... 11 Architecture and Design of VMware View 5.1 and NexentaVSA for View on Cisco UCS and Storage Solution... 11 Overview of Reference Architecture for a Persistent Desktop Pool... 11 High Availability and Failover... 13 Storage Layer... 13 Connectivity Layer... 13 Host Layer... 14 Overview of Reference Architecture for a Low-Cost Desktop Pool... 14 Solution Components: Cisco, Storage, and VMware-Based Reference Architecture... 15 Solution Validation... 16 Testing Methodology and Results... 16 Testing Methodology and Success Criteria... 16 Testing Profile Characteristics... 16 Load Generation... 17 User Workload Simulation: Login VSI from Login Consultants... 17 About the Heavy Workload... 18 Metrics... 19 Success Criteria: Login VSI... 19 Application Response Time Measured by VSImax Dynamic... 19 Login VSI Test 1: VSImax Connected for 196 Desktops with Medium-Sized Workload... 20 [VSILauncher.ini] - Configuration File for Login VSI Test 1... 20 Results... 21 IOPS Measured by NexentaVSA for View Management Appliance... 22 Login VSI Test 2: Cisco Recommended Number of Desktops for Medium-Sized Workload on Cisco UCS B230 M2... 24 [VSILauncher.ini] - Configuration File for Login VSI Test 2... 24 2013 Cisco Nexenta. All rights reserved. Page 2

Results... 24 Login VSI Test 3: Heavy Workload... 27 [VSILauncher.ini] - Configuration File for Login VSI Test 3... 27 Results... 27 Login VSI Test 4: Recommended Number of Desktops for Heavy Workload... 30 [VSILauncher.ini] - Configuration File... 30 VMware View Planner Testing... 33 Success Criteria: VMware View Planner... 33 VMware View Planner - Results File... 33 Results... 34 Success Criteria: IOmeter... 34 IOmeter - Results File... 35 Conclusion... 37 For More Information... 38 Appendix: Bill of Materials for NexentaVSA for View on Cisco UCS... 38 2013 Cisco Nexenta. All rights reserved. Page 3

Objectives This document describes the reference architecture of the Cisco Desktop Virtualization solution with storage for 500 to 1000 virtual desktops based on VMware View 5.1, VMware vsphere 5, and NexentaVSA for View (NV4V). The Cisco Desktop Virtualization solution includes the Cisco UCS B-Series Blade Servers and Cisco Nexus 5000 Series Switches. The purpose of the reference architecture is to provide tested and modular architecture built with proven best-inclass technologies to create a complete desktop virtualization solution that includes the desktop software, hypervisor, computing, networking, and storage elements and that is based on a virtual storage appliance approach integrated into the Cisco Unified Computing System (Cisco UCS) platform. This reference architecture accelerates your desktop transformation by enabling faster deployment, the flexibility of greater choice, increased efficiency, and lower risk. The modular architecture design uses the Cisco UCS B230 M2 Blade Server and transforms the Cisco UCS C240 M3 Rack Server into an enterprise storage appliance. Note that this reference architecture is not intended to be a comprehensive deployment and configuration guide for every aspect of this solution. Summary of Main Findings The combination of Cisco UCS, Cisco Nexus switches, and Nexenta virtual storage appliances with VMware ESXi 5 and VMware View software produces a virtual desktop delivery system with a high density per blade and chassis. Cisco maintains industry leadership with the new Cisco UCS Manager 2.0 software, which makes scaling simple, consistency essentially guaranteed, and maintenance easy. The Cisco 10 Gigabit Ethernet unified fabric is also validated on second-generation Cisco UCS 6200 Series Fabric Interconnects and second-generation Cisco Nexus 5500 platform access switches through more challenging workload testing, maintaining unsurpassed user-response times. Business Value Customers require a scalable, tiered, and highly available infrastructure on which to deploy their virtual desktop environments. Several new technologies are available to assist you in designing a virtual desktop solution, but you need to know how to use these technologies to get the most from your investment, support service-level agreements (SLAs), and reduce you total cost of ownership (TCO). This solution builds a replica of a common customer virtual desktop infrastructure (VDI) environment and validates the environment for performance, scalability, and capability. You achieve: Increased control and security of your global mobile desktop environment, which is typically your environment most at risk Better end-user productivity with a more consistent environment Simplified management, with the environment contained in the data center Better support for SLAs and compliance initiatives Lower operating and maintenance costs High availability and low capital costs using the same footprint without affecting TCO and while maintaining the performance needed to meet SLA requirements 2013 Cisco Nexenta. All rights reserved. Page 4

Modular Virtual Desktop Infrastructure Overview Cisco Data Center Infrastructure for Desktop Virtualization Cisco focuses on three main elements to deliver the best desktop virtualization data center infrastructure: simplification, security, and scalability. The software combined with platform modularity provides a simplified, secure, and scalable desktop virtualization platform. Simplified Secure Scalable High per-server virtual desktop density Unified management providing common view of the platform Predefined, validated infrastructure Uniformity achieved through a single hardware platform that includes storage, computing, and networking resources Virtual appliance that manages the provisioning of desktops and simultaneously deploys storage to reduce VDI complexity. Virtual desktop-aware access and control policies Virtual desktop-aware networking and on-demand provisioning Segmentation and network security policies across the LAN and WAN Capability to linearly scale up to thousands of desktops in a single domain Rapid desktop provisioning through service profiles Low-latency, high-bandwidth network for delivery of virtual desktops and multimedia Virtual storage appliance that uses a local server for linear storage scalability based on demand Savings and Success The simplified, secure, scalable Cisco data center infrastructure solution for desktop virtualization saves time and money. It provides faster payback and ongoing savings (better return on investment [ROI] and lower TCO) with the industry s highest virtual desktop density per server, meaning that fewer servers are needed, reducing both capital expenditures (CapEx) and operating expenses (OpEx). For example, NexentaVSA for View reduces server and storage costs by implementing a virtual storage appliance (VSA) on the same hardware on which virtual desktops are deployed. Network infrastructure costs also are much lower, with fewer cables per server and fewer ports required, using the Cisco UCS architecture and unified fabric. The simplified deployment of Cisco UCS for desktop virtualization makes you more productive faster and enhances business agility. IT staff and end users are more productive more quickly, and the business can respond to new opportunities simply by deploying virtual desktops whenever and wherever needed. The highperformance Cisco platform and network deliver a near-native end-user experience, allowing users to be productive anytime, anywhere. 2013 Cisco Nexenta. All rights reserved. Page 5

Project Planning and Solution Sizing Sample Questions Now that we understand user groups, their applications and their data requirements. In planning your solution, you need to consider some project and solution sizing questions. Here are some general project questions that should be addressed at the outset: Has a VDI pilot plan been created based on business analysis of the organization s desktop groups, applications, and data? Is the necessary infrastructure and budget in place to run the pilot program? Are the skill sets required to run the VDI project available? If not, can you hire or contract for them? Do you have end-user experience performance metrics identified for each desktop subgroup? How will you measure success or failure? What are the future implications of success or failure? Here is a partial list of sizing questions that should be addressed for each user subgroup: What desktop OS is planned? Microsoft Windows 7 or Windows XP? Will the OS be 32-bit or 64-bit? How many virtual desktops will be deployed in the pilot project? In production? Using Microsoft Windows 7? How much memory is needed per target desktop group desktop? Do you have any multimedia, Adobe Flash, or graphics-intensive workloads? What is the endpoint graphics processing capability? Are any Citrix XenApp hosted applications planned? Are they packaged or installed? What is the storage configuration in the existing environment? Are sufficient I/O operations per second (IOPS) available for the write-intensive VDI workload? Will you have storage dedicated and tuned for VDI service? Does the desktop have a voice component? Is antivirus software a part of the image? Is user profile management (nonroaming profile-based management) part of the solution? What are your fault-tolerance, failover, and disaster-recovery plans? Do you have any additional desktop subgroup-specific questions? 2013 Cisco Nexenta. All rights reserved. Page 6

The Solution: A Unified, Pretested, and Validated Infrastructure with Cisco, Nexenta, and VMware-Based Reference Architecture Cisco s desktop virtualization solution binds together the three critical elements of an end-to-end deployment: the end user, the network, and the data center. It draws on Cisco s architectural advantage to provide a solution that supports a diversity of endpoint devices, extends pervasive security and policy management to each virtual desktop, and uses a new and innovative virtualization-optimized stateless server computing model (Cisco UCS). Base Components Computing platform with Cisco UCS includes: Cisco UCS 6200 Series Fabric Interconnects Cisco UCS 2200 Series Fabric Extenders Cisco UCS 5108 Blade Server Chassis Cisco UCS B230 M2 Blade Servers for virtual desktop hosting Cisco UCS B200 M2 Blade Servers for infrastructure Cisco Nexus 5500 platform switches Hypervisor: VMware ESXi 5 Virtual desktop connection broker: VMware View 5.1 Storage: NexentaVSA for View and NexentaStor Cisco Unified Computing System Cisco UCS is the first truly unified data center platform that combines industry-standard, x86-architecture blade and rack servers with networking and storage access into a single system. The main innovations in the platform include a standards-based unified network fabric, Cisco virtual interface card (VIC) support, and Cisco Extended Memory Technology. The system uses a wire-once architecture with a self-aware, self-integrating, intelligent infrastructure that eliminates the time-consuming, manual, error-prone assembly of components into systems. Cisco UCS B-Series Blade Servers provide a comprehensive line of 2- and 4-socket servers to deliver worldrecord-setting performance to a wide range of workloads. Based on the Intel Xeon processor E7 and E5 product families, these servers are excellent for virtualized and nonvirtualized applications. These servers: Reduce CapEx and OpEx with converged network fabrics and integrated systems management Deliver performance, versatility, and density without compromise Address a broad set of workloads, from IT and web infrastructure through distributed databases for both virtualized and nonvirtualized environments Increase IT staff productivity and business agility through just-in-time provisioning and mobility support for both virtualized and nonvirtualized environments 2013 Cisco Nexenta. All rights reserved. Page 7

Cisco Nexus 5000 Series Switches The Cisco Nexus 5000 Series delivers an innovative architecture to simplify data center transformation by enabling a high-performance, standards-based, multiprotocol, multipurpose, Ethernet-based fabric. These switches help consolidate separate LAN, SAN, and server cluster network environments into a single 10 Gigabit Ethernet fabric. This unification enables network consolidation and greater utilization of previously separate infrastructure and cabling, reducing by up to 50 percent the number of adapters and cables required and eliminating redundant switches. This infrastructure displacement also lowers power and cooling costs significantly. VMware vsphere 5 VMware vsphere 5 is the market-leading virtualization platform and is used in thousands of IT environments around the world. VMware vsphere 5 transforms a computer s physical resources by virtualizing the CPU, RAM, hard disk, and network controller. This transformation creates fully functional virtual desktops that run isolated and encapsulated operating systems and applications just like physical computers. The high-availability features of VMware vsphere 5 are coupled with VMware Distributed Resource Scheduler (DRS) and vmotion, which enable the transparent migration of virtual desktops from one VMware vsphere server to another with little or no degradation of the customer s experience. This reference architecture uses VMware vsphere Desktop Edition for deploying desktop virtualization. It provides the full range of VMware vsphere Enterprise Plus Edition features and functions, allowing customers to achieve scalability, high availability, and optimal performance for all of their desktop workloads. Also, VMware vsphere Desktop comes with unlimited vram entitlement. VMware vsphere Desktop Edition is intended for customers who want to purchase only VMware vsphere licenses to deploy desktop virtualization. VMware View 5 VMware View is a desktop virtualization solution that simplifies IT manageability and control while delivering a high- fidelity end-user experience across devices and networks. VMware View helps IT departments automate desktop and application management, reduce costs, and increase data security through centralization of the desktop environment. This centralization results in greater end-user freedom and increased control for IT departments. By encapsulating the operating systems, applications, and user data in isolated layers, IT departments can deliver a modern desktop. They can then deliver dynamic, elastic desktop cloud services, such as applications, unified communications, and 3D graphics for real-world productivity and greater business agility. Unlike other desktop virtualization products, VMware View is built on and tightly integrated with VMware vsphere, the industry-leading virtualization platform that allows customers to extend the value of the VMware infrastructure and enterprise-class features, such as high availability, disaster recovery, and business continuity. NexentaVSA for View NV4V, from Nexenta, combines Nexenta's third-generation network-attached storage (NAS) and SAN storage appliance with a powerful VDI deployment and management server. NV4V is designed to interoperate transparently with VMware vcenter and View servers. Based on the Z File System (ZFS) architecture, NV4V is designed to use local storage devices with comprehensive ZFS features and provide calibration and measurement. You can use NV4V to deploy a new desktop pool, reconfigure an existing pool provisioned by NV4V to enhance resource utilization, calibrate an existing pool provisioned by NV4V, and run a performance benchmark on an existing pool provisioned by NV4V. NV4V is designed from the foundation to absorb and hide the complexity of virtual desktop data center deployment and management. The product provides fully automated, GUI-driven VDI provisioning, recalibration of existing desktop pools, I/O acceleration, performance graphics and analytics, built-in snapshots, and replication. It is a massively scalable, ready-to-use (turnkey) VDI solution. 2013 Cisco Nexenta. All rights reserved. Page 8

NV4V includes the following components: NexentaVSA for View Management Appliance is the main management component of NexentaVSA for View. The NV4V Management Appliance must be installed on the VMware ESXi Server that supports the management components. This NV4V component provides the wizards that deploy a desktop pool, calibrate the number of virtual desktops as, well as the amount of memory for the existing desktop pool, and run the performance benchmark tests. NexentaStor VSA provides storage for the NexentaVSA for View virtual desktops through a NV4V vsphere plug-in, which communicates with VMware View and VMware vcenter to perform desktop virtual machine (DVM) provisioning and management. During the deployment of a desktop pool, NexentaStor VSA is automatically installed on each VMware ESXi Server that is dedicated to virtual desktop storage. NexentaVSA for View Server Agent manages the communication between NexentaVSA for View and the VMware components. The server agent is installed on the VMware View Connection Server. NexentaVSA for View Desktop Agent provides communication between NexentaVSA for View and virtual desktops. The desktop agent is installed on the virtual desktop template. NV4V provides the following main features: Deployment automation Real-time monitoring and analytics Benchmarking and calibration Complete storage appliance I/O acceleration Local snapshot Remote backup and restore Replicated high availability NAS VMware vsphere Storage APIs and Array Integration (VAAI) Solution Overview and Benefits This solution uses Cisco UCS, Cisco Nexus 5548UP Switches, and VMware vsphere 5 to provide resources for a VMware View 5.1 environment of Microsoft Windows 7 virtual desktops provisioned by VMware View Composer. Careful planning and design of the server, networking, and storage infrastructure for the VMware View environment is critical. T server infrastructure needs to be sized to handle both the density and scale of the desktop workload, the networking infrastructure needs to provisioned to handle bursts of data traffic, and the shared storage needs to be able to absorb large bursts of I/O traffic that occur during a workday. To provide cost-effective and predictable performance for VDI, the infrastructure must be able to: Support a high density of virtual desktops per server Scale linearly with an increase in the number of virtual desktops Support rapidly provisioned scale-out infrastructure 2013 Cisco Nexenta. All rights reserved. Page 9

Provide low latency and high bandwidth for the clustering, provisioning, and storage interconnect networks Handle the peak I/O load from the clients while keeping response time quick Solution Benefits The NV4V solution on Cisco UCS blade servers enables you to use the Cisco UCS B230 M2 Blade Server as a computing node and the Cisco UCS C240 M3 Rack Server as storage for any VDI environment. In addition, the NV4V Management Appliance simplifies the time-consuming and complicated desktop pool deployment process with a six-step wizard. With traditional storage systems, customers must manually calculate the required IOPS and then calibrate the storage appliance without having received any real-time data from the VDI. This theoretical calibration can lead to guessing that often results in higher TCO and added complexity. By deploying local storage on the same computing node on which the VDI machines are deployed, NexentaVSA for View reduces the performance barriers that result from network constraints, storage protocol limitations, and external storage bottlenecks. The advantages include: Improved scalability: You can expand the existing VDI infrastructure when necessary. When you add VMware ESXi servers to support an increasing number of virtual desktops, you can reconfigure a desktop pool to use additional resources without the need to re-create the desktop pool. Reduced network load: NexentaStor VSA eliminates the need to provision additional network cables and network interface cards (NICs) for storage traffic, because it is installed on the same hardware as the virtual desktops and uses the internal network for storage traffic. Use of ZFS features: NexentaVSA for View inherits and uses the following ZFS features of the NexentaStor storage architecture: Hybrid storage pool: The hybrid storage pool enables you to use the fast solid-state drives as cache or log devices in a storage pool. NexentaStor VSA can provision a ZFS-based pool as storage for virtual desktops. ZFS compression: Default ZFS compression both improves IOPS and reduces the volume of storage space required for desktop pools. ZFS snapshot: NV4V provides fully integrated snapshot, backup and restore, and replicated high availability (RHA) features that use the underlying ZFS capabilities, which include unlimited snapshots. NAS VAAI plug-in: The plug-in offloads Network File System (NFS) traffic between NexentaStor VSA and VMware ESXi. Automatic reconfiguration and resource balancing: NexentaVSA for View enables you to optimize computing, memory, and storage parameters of new or existing desktop pools; it also configures the desktop pool according to parameters that you specify. Benchmarking and tuning: NexentaVSA for View provides sophisticated tools that analyze performance and enable you to calibrate a desktop pool. Calibration benchmarks adjust the number of desktops in a desktop pool based on the number of IOPS that are required. High availability for persistent desktop pool: Data integrity is crucial in persistent desktop pool deployments. NexentaVSA for View provides software data mirroring on a block level that enables automatic failover if a disk failure occurs. 2013 Cisco Nexenta. All rights reserved. Page 10

Performance analyzer: NV4V provides a set of performance analyzer utilities that help you monitor desktop pool activity and identify any potential performance bottlenecks. The performance analyzer utilities include: Built-in IOmeter tool that monitors the performance of a single desktop in a desktop pool Real-time performance charts that provide continuously updated statistical information Architecture Overview Figure 1 provides a high-level overview of the solution architecture. Figure 1: High-Level Architecture Overview Architecture and Design of VMware View 5.1 and NexentaVSA for View on Cisco UCS and Storage Solution Nexenta and Cisco provide the following joint solutions for VDI deployments: NV4V solution on Cisco UCS blade servers for highly available desktops NV4V solution on Cisco UCS blade servers for a persistent or stateless desktop pool Overview of Reference Architecture for a Persistent Desktop Pool The reference architecture for the solution includes hardware and software components that provide a reliable, scalable, and cost-effective solution for deployment of a persistent desktop pool with high-availability storage. 2013 Cisco Nexenta. All rights reserved. Page 11

The software components include: VMware ESXi cluster for virtual desktops VMware ESXi cluster for management components, such as Microsoft Active Directory (AD) NV4V Management Appliance NexentaVSA for View, which provides a virtual storage appliance and management functions For a fully redundant solution, the hardware components include: Cisco UCS 5100 Series Blade Server Chassis with eight Cisco UCS B230 M2 Blade Servers for virtual desktops and management components Two Cisco UCS C240 M3 Rack Servers for storage Two Cisco Nexus 5548UP Switches for network access for the Cisco UCS C240 M3 chassis Two Cisco UCS 6248UP 48-Port Fabric Interconnects to provide network connectivity and management capabilities to all Cisco UCS B230 M2 Blade Servers Figure 2 shows a NV4V solution on a Cisco UCS blade server for a persistent desktop pool. The components that are highlighted and enclosed in the red outline represent the tested configuration. The entire configuration as noted depicts fully configured Cisco UCS B230 M2 Blade Servers with a full complement of supporting storage on the Cisco UCS C240 M3 Rack Servers. For this highly available storage for VDI, two Cisco UCS C240 M3 Rack Servers are required to support one or two Cisco UCS B230 M2 Blade Servers for up to a total of 310 desktops. Figure 2: NV4V Solution on Cisco UCS Blade Server for Highly Available Virtual Desktops 2013 Cisco Nexenta. All rights reserved. Page 12

In the NV4V solution on Cisco blade servers for highly available desktops, storage is mirrored between two Cisco UCS C240 M3 Rack Servers serving NFS to the Cisco UCS B230 M2 Blade Servers. It provides a redundant and highly available infrastructure that is essential for a persistent desktop pool. Highly Available Storage for a Persistent Desktop Pool Software mirroring of two Cisco UCS C240 M3 Rack Servers provides highly-available storage. If one of the Cisco UCS C240 M3 servers fails, the other server is automatically used as storage. No additional plug-ins or administrative actions are required for this function. Figure 3 shows disk software mirroring on block-level storage. Figure 3: Block-Level Software Mirroring High Availability and Failover The NV4V solution on Cisco UCS blade servers for a persistent desktop pool provides a highly available virtual desktop infrastructure. Each component is configured to provide a reliable and scalable solution for all layers. Storage Layer The reference architecture for the NV4V solution on Cisco UCS blade servers for a persistent desktop pool is designed to mirror the hybrid pools on the Cisco UCS C240 M3 Rack Servers that you use as storage. If a hardware failure occurs on one of the Cisco UCS C240 M3 Rack Servers, the second server continues to operate. The redundant configuration of the disks on the back end protects against data loss resulting from harddisk failures. Connectivity Layer The NV4V solution on Cisco hardware provides redundancy on the connectivity layer. It includes use of two Cisco UCS 6200 Series Fabric Interconnects and two Cisco Nexus 5500 platform switches. 2013 Cisco Nexenta. All rights reserved. Page 13

Host Layer The VMware ESXi hosts have extra power supplies and network connections that help to reduce the number of failures on VMware ESXi Server components. In addition, the high-availability feature is enabled on the VMware ESXi cluster, which helps promptly recover virtual desktops if a major host failure occurs. Overview of Reference Architecture for a Low-Cost Desktop Pool The reference architecture for a low-cost desktop pool solution includes hardware and software components that are similar to the components that are used for the NV4V solution on Cisco UCS blade servers for a persistent desktop pool. The NV4V solution on Cisco UCS blade servers for a low-cost desktop pool includes a single Cisco UCS C240 M3 Rack Server. When you deploy a stateless desktop pool, data integrity and disaster recovery are not primary concerns. Therefore, server mirroring is not needed for this solution. The software components include VMware ESXi clusters for virtual desktops and VMware ESXi clusters for management components such as Microsoft Active Directory, NV4V Management Appliance, and NexentaVSA for View, which provides a VSA as well as management functions. The hardware components include: Cisco UCS 5100 Series Blade Server Chassis with eight Cisco UCS B230 M2 Blade Servers for the virtual desktops and management components Cisco UCS C240 M3 Rack Server for storage One Cisco Nexus 5548UP Switch for network access One Cisco UCS 6248UP fabric interconnect to provide network connectivity and management capabilities to all Cisco UCS B230 M2 Blade Servers in a chassis One 10 Gigabit Ethernet Intel E10G42BFSR X520-SR2 network card for NFS traffic Figure 4 shows the NV4V solution on Cisco UCS blade server for a persistent or stateless desktop pool. The components that are highlighted and enclosed in the red outline represent the tested configuration. The entire configuration as noted depicts a fully configured Cisco UCS B230 M2 Blade chassis with a full complement of supporting storage on the Cisco UCS C240 M3 Rack servers. For this scalable configuration, one Cisco UCS C240 M3 Rack Server can support up to two Cisco UCS B230 M2 Blade Servers. The trade-off of this lower-cost solution is no storage, switch, or fiber interconnect fault tolerance. 2013 Cisco Nexenta. All rights reserved. Page 14

Figure 4: Lower-Cost NV4V Solution on Cisco UCS Blade Servers for a Persistent or Stateless Desktop Pool Solution Components: Cisco, Storage, and VMware-Based Reference Architecture Cisco s desktop virtualization solution binds the following critical elements of an end-to-end deployment: End user Network Data center 2013 Cisco Nexenta. All rights reserved. Page 15

It draws on Cisco s architectural advantage to provide a solution that supports a diversity of endpoint devices, extends pervasive security and policy management to each virtual desktop, and uses a new and innovative virtualization-optimized stateless server computing model: Cisco UCS. Solution Validation The solution validation includes implementation, testing, and analysis of the obtained results. The solution validation helps ensure that the reference architecture meets the requirements of the VDI environment. The following workload generator tools were used to validate the NexentaVSA for View solution on Cisco UCS: Login VSI View Planner IOmeter Testing Methodology and Results This section discusses the testing methodology, success criteria, and results. Testing Methodology and Success Criteria The testing results focused on the entire process of the virtual desktop lifecycle by capturing metrics during the desktop boot-up, user logon, and virtual desktop acquisition (also referred to as ramp-up); user workload execution (also referred to as steady state); and user logoff for the hosted VDI model under test. Test metrics were gathered from the hypervisor, virtual desktop, storage, and load generation software to assess the overall success of an individual test cycle. Each test cycle was not considered to have passed unless all the planned test users completed the ramp-up and steady-state phases (described in the following sections) and unless all metrics were within the permissible thresholds and noted as success criteria. Three completed test cycles were conducted for each hardware configuration, and results were found to be relatively consistent from one test to the next. Testing Profile Characteristics Table 1 provides profile characteristics for the testing environment that was used to validate this solution. Table 1: Profile Characteristics Profile Characteristic Value Number of virtual desktops 196 Virtual desktop OS Microsoft Windows 7 Enterprise (32-bit) CPUs per virtual desktop 1 virtual CPU (vcpu) Number of virtual desktops per CPU core 10 RAM per virtual desktop 1 GB Average storage available for each virtual desktop 2 GB Maximum IOPS running IOmeter tests 148.29 Number of data stores used to store linked clones 1 Number of data stores used to store replicas 1 2013 Cisco Nexenta. All rights reserved. Page 16

Profile Characteristic Number of deployed virtual desktops per data store Value 225 Disk and RAID types for data stores RAID 10 20 x 300-GB 15,000-rpm SAS 2 x 300-GB 15,000-rpm SAS for NexentaStor 1 x 200-GB SSD for log (ZIL) 1 x 200-GB SSD for cache (L2ARC) Number of VMware clusters 1 Number of VMware ESXi Servers in each cluster 1 Number of virtual desktops in each cluster 1000 Load Generation In each test environment, load generators were used to simulate a load with multiple users accessing the VMware View 5 environment and processing a typical end-user workflow. To generate load within the environment, an auxiliary software application was required to generate the end-user connection to the VMware View environment, to provide unique user credentials, to initiate the workload, and to evaluate the end-user experience. In the hosted VDI test environment, session launchers were used simulate a scenario in which multiple users make direct connections to VMware View 5. User Workload Simulation: Login VSI from Login Consultants A critical factor in validating a VMware View desktop deployment is identification of a real-world user workload that is easy for customers to replicate and is standardized across platforms to allow customers to realistically test the effects of a variety of worker tasks. To accurately represent a real-world user workload, a third-party tool from Login Consultants was used throughout the hosted VDI testing. The tool measures the in-session response time, providing an objective measure of the expected user experience for individual desktops throughout large-scale testing, including during login storms. The Login Virtual Session Indexer (Login VSI 3.6) methodology, designed for benchmarking server-based computing (SBC) and VDI environments, is completely platform and protocol independent and hence allows customers to easily replicate the testing results in their environments. Note: The testing described here used the tool to benchmark the VDI environment only. Login VSI calculates an index based on the number of simultaneous sessions that can be run on a single machine. It simulates a medium-sized user workload (also known as a knowledge worker workload) running generic applications such as Microsoft Office 2007 or 2010, Microsoft Internet Explorer (IE) 8 including Adobe Flash applets, and Adobe Acrobat Reader. Note: For the purposes of this test, applications were installed locally, not streamed or hosted on Citrix XenApp. 2013 Cisco Nexenta. All rights reserved. Page 17

To simulate real users, the scripted Login VSI session leaves multiple applications open at the same time. The medium-sized workload is the default workload in Login VSI and was used for this testing. MediumNoFlash is a workload based on the medium-sized workload with only the Adobe Flash components disabled. Here is a summary of the testing: The workload emulated the medium-sized workload of a knowledge worker using Microsoft Office and IE and PDF files. After a session was started, the medium-sized workload repeated every 12 minutes. During each loop, the response time was measured every two minutes. The medium-sized workload opened up to five applications simultaneously. The type rate was 160 milliseconds (ms) for each character. Approximately two minutes of idle time was included to simulate real-world users. Each loop opened and used: Microsoft Outlook 2007 or 2010: 10 messages were browsed. Microsoft IE: One instance was left open (BBC.co.uk), and one instance was browsed to Wired.com, Lonelyplanet.com, and the heavy 480p Adobe Flash application gettheglass.com (not used with MediumNoFlash workload). Microsoft Word 2007 or 2010: One instance was used to measure response time, and one instance was opened to review and edit a document. Bullzip PDF Printer and Acrobat Reader: The Microsoft Word document was printed and reviewed as a PDF file. Microsoft Excel 2007 or 2010: A very large randomized sheet was opened. Microsoft PowerPoint 2007 or 2010: A presentation was reviewed and edited. 7-Zip: Using the command line, the output of the session was zipped. About the Heavy Workload The heavy workload required more memory and CPU consumption because additional applications ran in the background. The heavy workload had these characteristics: This workload simulated a power user. This workload was based on a medium-sized workload. The heavy workload differed from the medium-sized workload as follows: Type rate is 130 ms per character. Idle time total is only 40 seconds. The heavy workload opens up to eight applications simultaneously 2013 Cisco Nexenta. All rights reserved. Page 18

Metrics Multiple metrics were captured during each test run, but the success criteria for considering a single test run as pass or fail was based on these metrics: Login VSI Max, VMware View Planer, and IOmeter. The Login VSI Max metric evaluates the user response time during increasing user load and assesses the successful start-to-finish processing of all the initiated virtual desktop sessions. VMware View Planner is a VMware workload generator that you can use to simulate the workload in your VMware View environment. VMware View Planner is a tool that helps you determine the size and best characteristics of VDI deployments. It can be used as an alternative workload simulation tool to compare and validate the results of the Login VSI test. The VMware View Planner test results can be submitted to the VMware Rapid Desktop Program. IOmeter is a workload generator and a storage performance analyzer that you can use to measure the performance of a single desktop in a desktop pool. IOmeter is an industry-standard I/O subsystem measurement and characterization tool that is built into the NexentaVSA for View software. Success Criteria: Login VSI Application Response Time Measured by VSImax Dynamic The VSImax Dynamic test simulates the real-world VDI workload by constantly generating application loads that use computing and storage resources such as CPU, memory, disk capacity, and network bandwidth, through which the following operations generate the VSImax Dynamic score: Copying a new document from the document pool in the home drive Starting Microsoft Word with a document Opening the File Open dialog box Starting Notepad Opening the Print dialog box Opening the Search and Replace dialog box Compressing the document into a zip file using the 7-Zip command line These operations run simultaneously in the VSImax Dynamic test environment. To balance the operations and simulate a practical workload, you must properly weight the operations to reflect an accurate and practical test score. In this test, VSImax Dynamic with a 25-second logon interval was used to establish the benchmark of 196 desktops per Cisco UCS B230 M2 Blade Server. VSImax Dynamic takes a sample baseline measure of the Login VSI sessions and then performs a calculation to determine when the dynamic response-time threshold is reached, representing the density of the 196 desktops in the test that the server can support without degrading the user experience. In the baseline measures, an average response time of less than 4000 ms (4 seconds) was required for a passing test run. The NexentaVSA for View configuration achieved a 1264-ms baseline response time. In practice, the baseline response time is typically 1400 to 1800 ms without application virtualization or antivirus software. Table 2 and Figure 5 show the application response time by VSImax Dynamic for NexentaVSA for View with Cisco UCS B230 M2 and C240 M3 servers. 2013 Cisco Nexenta. All rights reserved. Page 19

For more information about Login VSI, please refer http://www.loginvsi.com/documentation/v3/calculating-vsimax. Table 2: Sample VSImax Dynamic Response-Time Calculation Activity Result (MS) Weight (Percent) Weight Result (MS) Refresh a document (RFS) 24 100% 24 Start Microsoft Word with a new document (LOAD) Launch the File Open dialog box (OPEN) 1487 33% 493 104 100% 104 Start Notepad (NOTEPAD) 33 300% 99 Launch the Print dialog box (PRINT) Launch the Replace dialog box (FIND) 44 200% 88 28 400% 112 Zip the document (ZIP) 172 200% 344 VSImax Dynamic Response Time 1264 Figure 5: VSImax Dynamic Activity Login VSI Test 1: VSImax Connected for 196 Desktops with Medium-Sized Workload Test 1 was a Login VSI test using a VSImax Connected parameter for 196 desktops with a medium-sized workload. [VSILauncher.ini] - Configuration File for Login VSI Test 1 [Launcher] Servername= Username= Password= Domain= CommandPassword= ConnectionType=Custom ConnectionNumber=User 2013 Cisco Nexenta. All rights reserved. Page 20

CCL=C:\Program Files\VMware\VMware View\Client\bin\wswc.exe -serverurl 10.0.101.72 - username Login_VSI%count% -password Q1w2e3r4! -domainname cisco -desktopname CiscoTest - Standalone -loginascurrent False -connectusbonstartup False -noninteractive /novmwareaddins CSV= Launchmode=Sequential PreTestScript= PostTestScript= ParallelDelay=10 ParallelTimeframe=1800 InitialStartNumber=1 NumberOfSessions=215 SequentialInterval=25 Fancy_number=1 Autologoff=1 LogoffTimeOut=900 CreateProfile=0 UseLocalLauncher=1 Results Figure 6 shows the results of a Login VSI test run for a medium-sized workload. Figure 6: Login VSI Test Results for Medium-Sized Workload Figure 7 shows the results of a Login VSI test that uses 196 virtual desktop during 2 hours with a medium-sized workload. The VSImax parameter is 196. The achieved Dynamic VSImax parameter is 4580 with a 1264 baseline, with an average read latency 0.639 ms (peak of 2 ms) and an average write latency 2.71 ms (peak of 6 ms). Figure 7 shows CPU utilization during the Login VSI test with a medium-sized workload. Figure 7: CPU Utilization During the Login VSI Test with a Medium-Sized Workload 2013 Cisco Nexenta. All rights reserved. Page 21

Figure 8 shows the data store read/write latency during the Login VSI test with a medium-sized workload. Figure 8: Data Store (Read/Write Latency) During the Login VSI Test with a Medium-Sized Workload for 196 Desktops Figure 8 shows the data store latency during the Login VSI test that uses 196 virtual desktop during a 2-hour period with a medium-sized workload. The achieved read latency was 0.639 ms and the achieved write latency was 2.711 ms, which is better than the average enterprise-class storage read latency results of 10 to 15 ms. IOPS Measured by NexentaVSA for View Management Appliance With traditional storage systems, customers have to manually calculate the required IOPS and then calibrate their storage appliances without any real-time data from their VDI infrastructure. This theoretical calibration can lead to guessing, often resulting in higher TCO and added complexity. By deploying local storage on the Cisco UCS B230 M2 and C240 M3 servers on which the VDI machines are deployed, the integration of NexentaVSA for View on the Cisco UCS platform enables rapid desktop provisioning using the NV4V Management Appliance. The NV4V Management Appliance can deploy virtual desktops and associate storage simultaneously from a single management console. In this case, the joint solution eliminates performance barriers to provide 148.29 IOPS per user. The NV4V Management Appliance runs Microsoft SQLIO and IOmeter to measure IOPS for a selected desktop pool. The IOPS results are displayed in real time through the NexentaVSA for View Management Appliance GUI. Testing parameters are set at a rigorous standard of 75 percent write operations and 25 percent read operations for 196 virtual desktops achieved; standard-industry practice is 75 percent read operations and 25 percent write operations. 2013 Cisco Nexenta. All rights reserved. Page 22

Figure 9 shows the NV4V IOmeter test results. Figure 9: IOPS for 196 Desktops from NV4V IOmeter Benchmarking Tool Figure 9 shows the results of a NV4V IOmeter benchmark test that uses 196 virtual desktops. The achieved results were 148.29 IOPS, with 25 percent read operations and 75 percent write operations. Figure 10 shows the NV4V Bootstorm test results. Figure 10: Bootstorm for 196 Desktops from NV4V Bootstorm Benchmarking Tool 2013 Cisco Nexenta. All rights reserved. Page 23

Figure 10 shows the results of a NV4V Bootstorm benchmark test that uses 196 virtual desktops. The achieved boot time is 1.71 seconds per desktop, with a total bootstorm 335 seconds. Login VSI Test 2: Cisco Recommended Number of Desktops for Medium-Sized Workload on Cisco UCS B230 M2 Test 2 was a Login VSI test with 155 desktops with approximately 90 percent CPU utilization. [VSILauncher.ini] - Configuration File for Login VSI Test 2 Servername= Username= Password= Domain= CommandPassword= ConnectionType=Custom ConnectionNumber=User CCL=C:\Program Files\VMware\VMware View\Client\bin\wswc.exe -serverurl 10.0.101.72 - username Login_VSI%count% -password Q1w2e3r4! -domainname cisco -desktopname CiscoTest - Standalone -loginascurrent False -connectusbonstartup False -noninteractive /novmwareaddins Launchmode=Parallel PreTestScript= PostTestScript= ParallelDelay=10 ParallelTimeframe=1800 InitialStartNumber=1 NumberOfSessions=155 SequentialInterval=25 Fancy_number=1 Autologoff=1 LogoffTimeOut=900 CreateProfile=0 UseLocalLauncher=1 Results Figure 11 shows the results of a Login VSI test run for a medium-sized workload with approximately 90 percent CPU utilization. Figure 11: Login VSI Test Results for Recommended Medium-Sized Workload with 155 Desktops 2013 Cisco Nexenta. All rights reserved. Page 24

Figure 11 shows the results of a Login VSI test that uses 155 virtual desktops during 1 hour with a medium-sized workload. The results showed an achieved 1264 baseline with an average read latency of 1.37 ms (peak of 4 ms) and an average write latency of 2.48 ms (peak of 14 ms). The VSImax value was not reached due to the low desktop count and low CPU utilization (approximately 90 percent). Figure 12 shows CPU utilization during the Login VSI test with a medium-sized workload. Figure 12: CPU Utilization During the Login VSI Test with a Medium-Sized Workload for 155 Desktops Figure 13 shows the data store read/write latency during the Login VSI test with a medium-sized workload. Figure 13: Data Store (Read/Write Latency) During the Login VSI Test with a Medium-Sized Workload for 155 Desktops Figure 13 shows the data store latency during a Login VSI test that uses 155 virtual desktop during a 2-hour period with a medium-sized workload. The achieved read latency is 1.372 ms and the achieved write latency is 2.483 ms, which is better than the average enterprise-class storage read latency of 10 to 15 ms. 2013 Cisco Nexenta. All rights reserved. Page 25

Figure 14 shows the results of the NV4V IOmeter benchmark test. Figure 14: IOPS for 155 Desktops from NV4V IOmeter Benchmarking Tool Figure 14 shows the results of a NV4V IOmeter benchmark test that uses 155 virtual desktops. The achieved result is 220.43 IOPS with 25 percent read operations and 75 percent write operations. Figure 15 shows the NV4V Bootstorm test results. Figure 15: Bootstorm for 155 Desktops from NV4V Bootstorm Benchmarking Tool Figure 15 shows the results of a NV4V Bootstorm benchmark test that uses 155 virtual desktops. The achieved boot time is 2.05 seconds per desktop with a total bootstorm of 318 seconds. 2013 Cisco Nexenta. All rights reserved. Page 26

Login VSI Test 3: Heavy Workload Test 3 was a Login VSI test with 168 desktops with a heavy workload. [VSILauncher.ini] - Configuration File for Login VSI Test 3 Servername= Username= Password= Domain= CommandPassword= ConnectionType=Custom ConnectionNumber=User CCL=C:\Program Files\VMware\VMware View\Client\bin\wswc.exe -serverurl 10.0.101.72 - username Login_VSI%count% -password Q1w2e3r4! -domainname cisco -desktopname CiscoTest - Standalone -loginascurrent False -connectusbonstartup False -noninteractive /novmwareaddins Launchmode=Parallel PreTestScript= PostTestScript= ParallelDelay=10 ParallelTimeframe=10 InitialStartNumber=1 NumberOfSessions=200 SequentialInterval=25 Fancy_number=1 Autologoff=1 LogoffTimeOut=900 CreateProfile=0 UseLocalLauncher=1 Results Figure 16 shows the Login VSI test for a heavy workload. Figure 16: VSImax Test Results for a Heavy Workload Figure 16 shows the results of a VSImax test that uses 195 virtual desktops during a 2-hour period with a heavy workload. The VSImax parameter is 168. The achieved Dynamic VSImax parameter value is 5153 with a 1722 baseline, with an average read latency of 0.69 ms (peak of 4 ms) and an average write latency 3.34 ms (peak of 19 ms). 2013 Cisco Nexenta. All rights reserved. Page 27

Figure 17 shows CPU utilization during the Login VSI test with a heavy workload. Figure 17: CPU Utilization During a Login VSI Test Run with a Heavy Workload for 168 Desktops Figure 18 shows the data store read/write latency during the Login VSI test with a heavy workload. Figure 18: Data Store (Read/Write Latency) During the Login VSI Test with a Heavy Workload for 168 Desktops Figure 18 shows the data store latency during the Login VSI test that uses 168 virtual desktops during a 2-hour period with a heavy workload. The achieved read latency is 0.694 ms and the achieved write latency is 3.339 ms, which is better than the average enterprise-class storage read latency of 10 to 15 ms. 2013 Cisco Nexenta. All rights reserved. Page 28

Figure 19 shows the NV4V IOmeter test results. Figure 19: IOPS for 168 Desktops from NV4V IOmeter Figure 19 shows the results of a NV4V IOmeter benchmark test that uses 168 virtual desktops. The achieved result is 178.26 IOPS with 25 percent read operations and 75 percent write operations. Figure 20 shows the results of the NV4V Bootstorm benchmark test. Figure 20: Bootstorm for 168 Desktops Figure 20 shows the results of a NV4V Bootstorm benchmark test that uses 168 virtual desktops. The achieved boot time is 1.93 seconds per desktop with a total bootstorm of 325 seconds. 2013 Cisco Nexenta. All rights reserved. Page 29

Login VSI Test 4: Recommended Number of Desktops for Heavy Workload Test 4 was a Login VSI test with 125 desktops with a heavy workload. [VSILauncher.ini] - Configuration File Servername= Username= Password= Domain= CommandPassword= ConnectionType=Custom ConnectionNumber=User CCL=C:\Program Files\VMware\VMware View\Client\bin\wswc.exe -serverurl 10.0.101.72 - username Login_VSI%count% -password Q1w2e3r4! -domainname cisco -desktopname CiscoTest - Standalone -loginascurrent False -connectusbonstartup False -noninteractive /novmwareaddins Launchmode=Parallel PreTestScript= PostTestScript= ParallelDelay=10 ParallelTimeframe=1800 InitialStartNumber=1 NumberOfSessions=125 SequentialInterval=25 Fancy_number=1 Autologoff=1 LogoffTimeOut=900 CreateProfile=0 UseLocalLauncher=1 Figure 21 shows the results of the test. Figure 21: Login VSI Test Results for Recommended Heavy Workload with 125 Desktops Figure 21 shows the results of a Login VSI test that uses 125 virtual desktop during a 1-hour period with a heavy workload. The results showed an achieved baseline of 1264 with an average read latency of 1.40 ms (peak of 3 ms) and an average write latency of 2.19 ms (peak of 5 ms). 2013 Cisco Nexenta. All rights reserved. Page 30

Figure 22 shows CPU utilization during the Login VSI test with a heavy workload. Figure 22: CPU Utilization During the Login VSI Test with a Heavy Workload for 125 Desktops Figure 23 shows the data store read/write latency during the Login VSI test with a heavy workload. Figure 23: Data Store (Read/Write Latency) During the Login VSI Test with a Heavy Workload for 125 Desktops Figure 23 shows the data store latency during a Login VSI test that uses 125 virtual desktop during a 2-hour period with a heavy workload. The achieved read latency is 1.397 ms and the achieved write latency is 2.192 ms, which is better than the average enterprise-class storage read latency of 10 to 15 ms. 2013 Cisco Nexenta. All rights reserved. Page 31

Figure 24 shows the NV4V IOmeter test results. Figure 24: IOPS for 125 Desktops from NV4V IOmeter Figure 24 shows the results of a NV4V IOmeter benchmark test that uses 125 virtual desktops. The achieved result is 270.76 IOPS with 25 percent read operations and 75 percent write operations. Figure 25 shows the results of the NV4V Bootstorm benchmark test. Figure 25: Bootstorm Results for 120 Desktops Figure 25 shows the results of a NV4V Bootstorm benchmark test that uses 125 virtual desktops. The achieved boot time is 1.92 seconds per desktop with a total bootstorm of 241 seconds. 2013 Cisco Nexenta. All rights reserved. Page 32

VMware View Planner Testing To simulate a medium-sized user workload, the VMware View Planner runs the following applications: Adobe Reader Microsoft Excel Mozilla Firefox Microsoft Internet Explorer Microsoft Outlook Microsoft PowerPoint Microsoft Word Video applications 7-Zip Success Criteria: VMware View Planner The success of the VMware View Planner test run is determined by the VMware View Planner score. A VMware View Planner score represents the number of concurrent virtual desktops that participate in a successful test run. A successful test run must meet the following requirements: It must run all the standard VMware View Planner applications and no other applications. The think time must be set to 20 seconds. It must include at least five iterations (including ramp-up and ramp-down). At least 95 percent of the application response times during steady-state activity must be 1.5 seconds or less. VMware View Planner - Results File Test Name: CiscoTest2 Test Mode: local QoS Summary ----------- Group A : PASSED The 95th percentile was: 0.742654 seconds (To pass, this must not be more than 1.5 seconds) Workload Summary ----------------- Users: 200 Iterations (Total): 5 (to pass, this must be at least 5) Iterations (Scored): 3 Workload Status: PASSED 2013 Cisco Nexenta. All rights reserved. Page 33

Results Figure 26 shows the results from the VMware View Planner CPU test. Figure 26: Results from VMware View Planner CPU Test Figure 26 shows the results from a VMware View Planner CPU test that ran using 200 desktops during a 6-hour period with a medium-sized workload. The 95th percentile value for the application response time was 0.742654 second, which is below the defined threshold of 1.5 seconds. Success Criteria: IOmeter Figure 27 shows different types of VDI users with their storage performance requirements and IOPS. Figure 27: Types of VDI Users and Performance Requirements According to third-party studies, the expected number of IOPS for task users is 4 to 6, for knowledge workers it is 7 to 12, and for power users it is 25 to 50; the average number of IOPS for mobile users is 20. 2013 Cisco Nexenta. All rights reserved. Page 34