HP Client Virtualization Reference Architecture for HP VirtualSystem

Transcription

1 HP Client Virtualization Reference Architecture for HP VirtualSystem Virtual Desktop Infrastructure for the Enterprise Technical white paper Table of contents HP and Client Virtualization... 2 The many ways to virtualize a client... 3 Building a successful Client Virtualization model with VDI and HP... 4 User virtualization... 5 Application virtualization... 5 Core compute resource... 6 User devices... 6 Device management... 6 Services... 6 Data center infrastructure... 6 Understanding the VDI Client Virtualization Reference Architecture... 8 Reference Architecture data center infrastructure building blocks HP BladeSystem HP Virtual Connect Flex HP ProLiant servers HP storage Management Access devices HP Thin Clients HP Services Just beginning your Client Virtualization journey? Planning to implement Client Virtualization, but want the shortest path to the best plan? Starting a Windows 7 migration and investigating Client Virtualization? Want HP to design and build your solution, and turn it over to you? Want HP to design and manage your client infrastructure solution? Constructing the Reference Architecture Enterprise Including direct attached storage End users and server performance About end users Configuration of recommended platforms HP test methods and results Performance factors outside of user type HP Client Virtualization Analysis and Modeling service For more information Call to action... 41

2 HP and Client Virtualization Planning a Microsoft Windows 7 migration? How much of your corporate data is at the airport today in a lost or stolen laptop? What is the cost per year to manage your desktops? Are you prepared to support the upcoming always-on workforce? HP Client Virtualization can help customers achieve the goals of IT and workforce support, without compromising on performance, operating costs, information security, and user experience with HP Client Virtualization Reference Architectures. These reference architectures provide: Simplicity: with an integrated data center solution for rapid installation/startup and easy ongoing operations Self contained and modular server, storage, and networking architecture no virtualization data egresses the rack 3x improvement in IT productivity Optimization: a tested solution with the right combination of compute, storage, networking, and system management tuned for Client Virtualization efficiency Scalable performance, enhanced security, always available 60% less rack space compared to competitors 95% fewer NICs, HBAs, and switches; 65% lower cost; 40% less power for LAN/SAN connections Flexibility: with options to scale up and/or scale out to meet precise customer requirements Flexible solution for task workers to PC power users Combine up to 10,500 productivity and task workers in three racks Unmatched price/performance with both direct attached (DAS) and SAS tiered storage in a single rack (up 50% cheaper than traditional fibre channel SAN) The complete reference architecture is a tool for HP VirtualSystem, a strategic portfolio of infrastructure solutions, which serves as the foundation for your virtualized workloads. Based on HP Converged Infrastructure (CI), HP VirtualSystem utilizes market-leading capabilities to centralize administrative tasks, improve scalability, optimize workloads and reduce complexity. Target audience This document is targeted at IT decision makers and architects that are looking for guidance to construct a complete integrated Client Virtualization solution. This document is intended as an overview document and is accompanied by additional reference architecture documents highlighting integrated solutions with major virtualization vendors, available at 2

3 The many ways to virtualize a client While this document focuses on building out a platform for desktop virtualization, there are other approaches to supplying end users with an experience that meet the goals for lower costs, simpler management and a more flexible infrastructure. In many cases, these approaches can and even should be combined with virtual desktop infrastructure (VDI) to create the most flexible infrastructure available. These approaches include: Application virtualization Server based computing Blade workstations Cloud services Application virtualization With application virtualization, the application itself is decoupled and managed separately from the core operating system. There are a variety of methods to do this including application streaming and application packaging. Under ideal circumstances, you should be able to use the best tool or combination of tools to meet your needs without incurring extra management headaches or having to force your users into a single use model that may not meet all of your needs. Aside from being a standalone approach to providing an end user experience, application virtualization should be considered essential to a successful VDI implementation. Server based computing Server based computing has a long and rich history of satisfying the computing needs of a variety of users. With server based computing, multiple desktops are presented from a single Microsoft Windows host. This generally leads to very high consolidation ratios and centralized management for users that have a light compute footprint. Blade workstations HP has long enjoyed a strong market position deploying Client Virtualization to end users that require a robust compute environment. These users may include graphics designers, financial traders or any number of heavy compute users. An HP ProLiant WS460c Workstation Blade provides a local workstation experience to end users over the network using the HP Remote Graphics Software protocol, Citrix HDX 3D and Microsoft Remote Desktop Protocol 7.1. Cloud services Whether applications delivered as a service or entire desktops delivered as a service, the cloud is coming. This approach can be inclusive of and/or complementary to VDI and should be comprehended as a part of a long term strategy for desktop alternatives. With each of these Client Virtualization approaches, user data is moved to a central location (or numerous central locations), control of compute resources is centralized, compute resources are standardized resulting in a homogenous management target for IT and support, a remote protocol is used to access a compute resource and the choice of access device has grown exponentially. 3

4 Building a successful Client Virtualization model with VDI and HP A successful architecture that meets the goals for Client Virtualization must include building blocks that ensure each layer is optimized from both a cost and manageability standpoint. Without these building blocks, it will be difficult to achieve the cost and flexibility goals that Client Virtualization implementations promise. The core building blocks, as shown in Figure 1, include user and application virtualization acting on a core compute resource. This resource is accessed via a user device. The core components are built on a robust data center infrastructure. Correctly selecting, designing and implementing each of these layers will create a cost effective, robust and highly flexible Client Virtualization environment. Figure 1: The HP Converged Infrastructure for Client Virtualization Reference Architecture for VDI HP Consulting, Integration and Factory Express Services User Devices Device Management User Virtualization Client Automation E-print Universal Print Drivers Compute Resource Application Virtualization Storage HP Converged Infrastructure Servers Power & cooling Network Management software HP Reference Architectures The following sections explain each of these layers. Check for other documents in the reference architecture series that address these areas. 4

5 User virtualization User virtualization in the enterprise involves the removal of the user s core personality from the compute resource. Traditionally, approaches like roaming profiles have been used to allow data segmentation which results in users being able to access more than one device while keeping their data accessible. Traditional approaches though have frequently required large amounts of network bandwidth, extended logon times and led to issues with cross operating system compatibility. To implement a cost effective VDI solution, these areas need to be addressed as well as insuring that the end user is not locked to any device in particular. The results of successfully implementing user virtualization frequently include the following capabilities: Administrators can place a device into the end user environment with little or no configuration and have a user become immediately productive. The end user can move between operating systems and devices without needing to reconfigure their environment. Administrators should be able to base the presentation of and access to resources on where a user is logged in as well as the device type and operating system. Changes to the user s environment should become available to the user immediately. The solution should be aware of the environment to allow for situational policy implementation. The compute resource being utilized should be able to be eliminated without removing user data and settings from the overall environment. Achieving these capabilities can be difficult, but when successfully implemented, they create an environment where it is far easier to virtualize the client in a number of ways as well as simple to alter the compute resource. User virtualization in an optimal form creates the ability to have end users accessing a device based on a Microsoft Windows XP Professional operating system leave for the evening and come back the next workday to a Microsoft Windows 7 Professional based device with their data, desktop settings, application customizations and overall experience in place. HP recommends virtualizing the user settings in all Client Virtualization implementations. Application virtualization Application virtualization refers to a set of technologies that disaggregates the application from the core compute resource and end user. An application that is virtualized may be packaged as an executable and run directly in the virtual machine (VM), installed and run on a remote system with the end user experience delivered via a display protocol, or divided between the core compute resource and an external server where pieces of the application are delivered locally only when needed. Each of these approaches carries merits and can be used across numerous solutions from VDI to physical desktops to potentially even tablet and phone devices. When used properly, one potential benefit of application virtualization is the removal of unique application information from the core compute resource. Once removed, the core compute resource becomes more flexible and, ideally, even disposable. Another potential benefit is the centralization of application updates to a single location. Of further benefit, application virtualization can also help serve legacy and troublesome applications to multiple, varying endpoints with minimal administrative involvement. For maximum effectiveness, greatest degree of solution flexibility and best return on investment, HP recommends the virtualization of applications in all VDI environments. 5

6 Core compute resource When properly architected the core compute resource in Client Virtualization becomes a simple to manage shell that is easy to change and ready to adapt. The goal is to manage as little as possible in this layer while still providing end users with the experience they require for their job. This resource could be a desktop, a thin client, or, as in the case of VDI, a virtual machine that is hosted and under the control of IT. This layer is not only the compute resource, but also the software required to create, manage and secure that resource from end-to-end. This includes insuring only users you want accessing the resource can access it, only the data you want displayed on the resource is displayed and only the programs you want delivered to it are delivered. HP partners with the leading vendors in the Client Virtualization space to insure end to end enterprise solutions that are tested and supportable and that provide a great end user experience. Check the partner reference architectures for Client Virtualization at User devices This layer includes all of the potential devices that an end user can bring to the compute environment. In simplest terms this includes the client itself as well as the printing infrastructure. In more complex environments this device count can of course grow quickly. Portable hard drives, USB keys, card readers and other security devices, tablets and smart phones can quickly pervade. HP offers tremendous value with this layer by offering not only access devices, but also the software to manage these devices as a core part of a more cohesive client infrastructure. Device management What you access the core compute resource with matters and the ability to manage those devices as part of a core device management strategy is critical. IT is now faced with more devices than ever. Many of these devices, while bringing value to end users, bring unique challenges to IT. Managing different devices, including VMs, in a different way from each other results in IT silos that are tough to converge later not to mention expensive to manage now. HP brings Client Automation to the data center to insure that the greatest number of devices can be managed with a single, sweeping approach. Services For customers adopting Client Virtualization, HP offers a valuable range of Client Infrastructure Services to help customers reduce TCO, increase security, readily adapt to new business requirements and simplify management. These services complement the HP Client Virtualization Reference Architectures with planning, design, and integration services for a complete solution. Engage with HP at your point of need, and get the business outcomes desired. Since no organization has the same goals, issues, and opportunities as other organizations, HP services are tailored to provide what you need to accelerate the planned results. Data center infrastructure At the core of every enterprise VDI implementation is a set of data center infrastructure that is optimized to work together, simple to manage and that can react to changing business needs. HP s Converged Infrastructure approach to solving VDI issues is unique in the industry in the level of integration it brings and the benefits that result. 6

7 HP has built its enterprise architecture on a series of building blocks that can be implemented to address specific use cases. While these use cases are discussed in depth later in the document, Figure 2 presents them at a high level. Figure 2: The HP platforms for the enterprise Converged Infrastructure reference architecture HP s enterprise configuration is built on HP BladeSystem featuring ProLiant servers, Virtual Connect Flex-10 and either HP P4800 G2 SAN Solution for BladeSystem or HP 600 Modular Disk System (MDS600). This configuration brings high levels of convergence to the server, storage and networking space resulting in simpler management as well as outstanding performance. A single administrator controls the entire hardware stack resulting in quick changes in reaction to business requirements. 7

8 Understanding the VDI Client Virtualization Reference Architecture VDI is one possible implementation of the Client Virtualization (CV) Reference Architecture. Blade Workstations and server based computing also fit the CV model. With VDI, a desktop is created as a virtual machine. Applications and user personality are injected into the core desktop virtual machine and a brokering mechanism manages connecting end users to the VM via a connection protocol. Figure 3 shows the architecture of a VDI implementation. Figure 3: Architectural overview of VDI More than simply a virtual machine, the runtime VDI instance is the real time compilation of the end user s data, personal settings and application settings with a core operating system instance and shared generic profile where the applications are either installed locally as a fully installed or packaged instance or streamed from outside the VM. While seemingly complex at first glance, the proper management of these individual pieces yields a far more flexible end user computing environment that is simpler to manage. This is accomplished in part by the decoupling of the user and application from the core operating system. The single OS image with few applications is easy to manage as part of an overall, comprehensive client management strategy using tools such as HP Client Automation. Once the user and application are segmented from the device, the device itself starts to matter less as well. With the right approach, an end user can receive the same experience regardless of what core compute resource they are logged onto at the moment. This also means that changing the OS layer is greatly simplified making tasks like Windows operating system migrations simpler for users who have been virtualized. Figure 4 highlights the compute device at runtime in more detail. 8

9 Figure 4: The VDI runtime instance This computing resource becomes available to end users in a variety of locations including branch offices, work from home and outsourced workers. Local users are of course a given. The compute resource thus becomes another resource that resides on the network core. Figure 5 highlights the overall architecture at a high level. Figure 5: VDI within the overall enterprise 9

10 Figure 6 shows the architecture from the network perspective. There are four networks that must be considered when utilizing the HP architecture. The first is the Internet itself which enables off campus users to access a compute resource. The second is the production WAN/LAN to enable corporate users to achieve an end user experience. A management network is also in place to support the management of the hardware via a variety of mechanisms including HP Integrated Lights-Out (ilo). Finally, a storage network is defined to allow communication between initiators and targets. One of the key differentiators for this architecture when compared against other offerings is the emphasis on network flexibility and cost for the core of the environment. This is accomplished by migrating as much of the infrastructure and compute resource as possible to within the Virtual Connect domain. This eliminates or minimizes the need for switch ports, allows for the management of intra-domain traffic by a single infrastructure administrator and improves performance and reliability by utilizing mostly cable free internal connections between hosts and management services. Figure 6: The HP Client Virtualization Reference Architecture for VDI with an emphasis on networking From an HP standpoint, the value of VDI is derived from the fact that HP provides an end-to-end solution. This simplifies support post sale, insures integrated management and function between components and gives you a stable platform that will never have its value determined by the ability of any one particular vendor among many to keep up with technology changes. If we look at the previous figure (Figure 6), it is clear that Virtual Connect plays a critical role in the overall architecture. Figure 7 is a closer view of how traffic flows and is segmented between the 10

11 server and the core. Of particular note is the minimized number of uplinks. A Virtual Connect domain is cabled directly to the core with a minimal number of cables while maintaining required bandwidth and availability. This eliminates the need for rack switches which not only saves cost, but also greatly simplifies the management of the network. In the figure below, production traffic and management traffic that need to reach the network core are sent via shared uplinks. Of equal interest are the internal networks created. These networks, labeled storage and special function in the figure below, traverse between enclosures but never egress the Virtual Connect domain and thus never touch core network infrastructure. In a VDI environment where numerous teams may need to interact simply to keep the environment up and running it can be difficult to manage the various levels of interactions. The internal storage and special function networks can reduce or even eliminate involvement from the storage and network teams. This in turn centralizes ownership of the solution stack and increases flexibility. One final note is that every ProLiant server blade is equipped with at least two (2) onboard 10Gb Flex-NICs each allowing for the creation of up to four (4) redundant networks of variable bandwidth between 100Mb and 10Gb. Figure 7: Networks internal and external to the Virtual Connect domain 11

12 In the context of VDI, there is an additional layer that must be comprehended. Within VDI environments, each hypervisor host communicates through an abstracted virtual networking layer that allows the individual VDI and management VMs to communicate with the production network as well as the individual hypervisor hosts to transmit traffic to storage, be managed and make use of any special function networks. Figure 8 highlights the relationship between the server, hypervisor, virtual machines, virtual networking layer, Flex-NICs and Virtual Connect. Figure 8: Virtual networking stack from Flex-NIC through to the virtual machine For specific information about implementing VDI on the HP Client Virtualization Reference Architecture using specific HP partner solutions, consult the HP partner reference architectures at 12

13 Reference Architecture data center infrastructure building blocks The HP Client Virtualization Reference Architecture is built on HP s award winning data center platforms. Incorporating servers, storage and networking to create a true Converged Infrastructure yields a flexible, cost effective platform to deploy and manage VDI. This section outlines the components of the platform and highlights the value that each brings to a VDI implementation. HP BladeSystem Drive business innovation and eliminate sprawl with HP BladeSystem, the industry s only Converged Infrastructure architected for any workload from client to cloud. HP BladeSystem is engineered to maximize every hour, watt, and dollar, saving up to 56% total cost of ownership over traditional infrastructures 1. With BladeSystem, it is possible to create a change ready, power efficient, network optimized, simple to manage and high performance infrastructure on which to build and scale your VDI implementation. An HP BladeSystem c7000 enclosure populated with HP ProLiant BL460c G7 blades is shown in Figure 9. Figure 9: The HP BladeSystem c7000 enclosure (pictured with HP ProLiant BL460c G7 server blades) Benefits for VDI HP BladeSystem allows for the use of high scaling, distributed 2-socket servers in VDI implementations while maintaining a minimal amount of infrastructure to be managed. The result is a highly available, 1 13

14 power efficient, simple to manage infrastructure set that sacrifices nothing while delivering optimized costs, converged storage and networking, and simple scalability. HP Virtual Connect Flex-10 HP VDI and Flex-10 technology HP Virtual Connect Flex-10 technology creates a dynamically scalable internal network architecture for VDI deployment. Multiple c7000 enclosures make up the core building blocks of the physical infrastructure. Each enclosure contains two Virtual Connect (VC) Flex-10 interconnect modules. Each module connects to two dual port 10Gb Flex-10 or FlexFabric adapters in each server (G7 BladeSystem servers have embedded FlexFabric adapters and can accommodate mezzanine Flex-10 or FlexFabric adapters). Each Flex-10 or FlexFabric adapter has four FlexNICs on each of its dual ports. The multiple FlexNICs can support the iscsi storage, specialized virtual machine function, management, and production networks recommended for HP VDI. VC Flex-10 modules and adapters aggregate Ethernet and accelerated iscsi storage traffic between the server and Flex-10 module (server-network edge) into a 10Gb link. Flex-10 technology partitions the 10Gb data stream into multiple (up to four) adjustable bandwidths, preserving routing information for all data classes. Flex-10 modules and stacking You can link the Flex-10 modules within each enclosure, and between enclosures, using the internal links available in each enclosure, and using stacking cables and appropriately configured ports on each module. This multi-enclosure stacking creates a single Virtual Connect domain between all the enclosures (up to four enclosures). This means that all HP BladeSystem VDI servers and the P4800 SAN operate in the same VC domain. The ability to consolidate all VDI server/storage-side traffic on a single internal VC network, is a major advantage. Any network adapter can connect with any other adapter within the VC network (domain) without exiting to the external network, and without additional switches or network management. This simplifies management, reduces support requirements and a tremendous reduction in both uplink and switch ports. Figure 10 shows a Virtual Connect Flex-10 module. Figure 10: The HP Virtual Connect Flex-10 Ethernet module 14

15 Benefits for VDI Network isolation Flex-10 technology provides physical network isolation for required network types. VM traffic within the server has access to eight FlexNICs per dual port Flex-10 or FlexFabric adapter (four FlexNICs per port). Network aggregation The Flex-10 or FlexFabric adapter aggregates each set of four FlexNICs into a single 10Gb uplink so that each dual port adapter has only two 10Gb links between the server adapter and the Flex-10 module. From the enclosure or Virtual Connect domain to the core, as few as two cables aggregate networks you choose to egress straight to the core resulting in a dramatic reduction in cables as well as expensive switch ports. Single internal network You can stack Flex-10 modules within the enclosure and between enclosures to create a single internal VC network (domain), eliminating the need for additional switches and greater intervention by network administrators, and in the case of the P4800 G2 SAN Solutions for BladeSystem, storage administrators. Unified management The Onboard Administrator, Virtual Connect Enterprise Manager and P4000 Centralized Management Console reduce and simplify the management burden for VDI administrators. Management and monitoring of the core pieces can be surfaced into your virtualization management software to further simplify the management picture. HP ProLiant servers Choosing a server for VDI involves selecting a server that is the right mix of performance, price and power efficiency with the most optimal management. HP s experience during test and in production has been that 2 socket servers are the ideal platform for VDI. With the possibility of over 100 VMs running on one platform, 2 socket systems offer better memory performance and thus better scaling models than larger systems. HP BladeSystem reduces costs and simplifies management through shared infrastructure. HP ProLiant BL460c G7 The HP ProLiant BL460c G7 server blade offers an excellent balance of performance, scalability and expandability for any workload, making it the standard for dense data center computing. Figure 11: The HP ProLiant BL460c G7 15

16 The HP ProLiant BL460c G7 offers Intel Xeon 5600 series processors, embedded FlexFabric adapters supporting Flex-10 and Ethernet, iscsi and FCoE traffic, up to 384GB of DDR3 memory, two hot plug drive bays with an embedded HP Smart Array controller as well as industry leading ilo 3 management capabilities. Benefits for VDI: The BL460c G7 is an ideal platform for task workers and VMs with memory footprints of 1GB and below. The BL460c offers not only high user counts per host, but also incredible density which is far more important in optimizing power and management efficiency as well as reducing infrastructure. HP ProLiant BL490c G7 The BL490c G7 meets the needs of the data center to be scalable, flexible, efficient, and change ready. With features tailored for virtualization, the BL490c G7 provides high-density power-saving options, a full suite of network interconnectivity options, including 10Gb Ethernet with converged network support, scalable performance and memory, and efficient, automated management tools. Figure 12: The HP ProLiant BL490c G7 The HP ProLiant BL490c G7 offers Intel Xeon 5600 series processors, embedded FlexFabric adapters supporting Flex-10 and Ethernet, iscsi and FCoE traffic, up to 384GB of DDR3 memory, up to two onboard solid state disks and industry leading ilo 3 management capabilities. Benefits for VDI: The BL490c G7 is an ideal platform for productivity workers and VMs with memory footprint of 2GB and below. Like the BL460c G7, it maximizes user density as well as power and management efficiency but provides a larger memory capacity using smaller DIMMs. 16

17 HP ProLiant BL620c G7 The BL620c G7 is the world s most scalable x86 2-socket blade now supporting the Intel Xeon processor E and E families. The BL620c G7 is perfect for 2P applications requiring performance and features greater than typical 2P server offerings. The scalability of the BL620c G7 offers up to 800% the memory, an extensive list of RAS features, 20x the embedded Ethernet bandwidth, 4x the VMs, 3x the on-board I/O expansion slots, and 4.8x the total I/O bandwidth than many other 2P blades available today. Figure 13: The HP ProLiant BL620c G7 Benefits for VDI: The ProLiant BL620c offers extremely high memory counts along with a large cache, core dense CPU architecture that provides maximum user counts when working with VMs with memory footprints of 2GB or greater. While some other vendors require greater infrastructure investments to squeeze fewer, smaller cache, lower core count servers into a 42U rack, the ProLiant BL620c G7 yields greater density with a single infrastructure approach and minimal cabling to the core utilizing Virtual Connect domains. The BL620c is a great choice for a knowledge worker VDI platform. 17

18 HP storage Storage is as important as the servers in your Client Virtualization deployment. The storage infrastructure is directly in the critical path for hundreds or thousands of your end users and this introduces a number of unique challenges with respect to delivering consistently high performance, minimizing management and deployment costs, and scalability and efficient capacity utilization. For this reason, the storage platform you choose is critical to making the economics of Client Virtualization work while maintaining high quality of service levels to satisfy user requirements. HP provides storage solutions to meet small to large Client Virtualization deployments with a range of SAN, NAS and DAS solutions. HP P4000 SAN Solutions ideal storage for Client Virtualization HP P4000 SAN solutions are optimized for virtual server and desktop environments and are based on a clustered storage architecture that eliminates single points of failure and scales nondisruptively in capacity and performance. All P4000 SAN solutions come with all SAN software features included. HP P4000 SAN solutions come in rack, HP BladeSystem, and Virtual SAN Appliance (VSA) software versions and deliver: Efficiency: Thin provisioning raises storage utilization and efficiencies Clustering, density, and storage tiers optimize dollars per I/O Performance: Clustered, virtualized architecture delivers maximum disk utilization Dense spindle count, high-speed storage paths, optimized use of SSDs High availability: No single-point-of-failure storage architecture through Network RAID Nondisruptive software upgrades Scalability: Nondisruptively scale capacity and performance Avoid up-front costs and scale as needs grow For more information, see 18

19 HP P4800 G2 SAN Solutions for BladeSystem The HP P4800 SAN Solution for BladeSystem delivers storage that is fine-tuned for Client Virtualization. It is based on a tight integration with HP BladeSystem and offers a highly dense storage solution that dynamically scales as the infrastructure expands. Being converged into the HP BladeSystem architecture provides network convergence of the virtual server and SAN using 10Gig Ethernet and Flex-10 Virtual Connect, enabling simplified administration, improved security, flexibility, and enhanced performance. Simplicity and performance are enhanced by no external storage switching, high-speed storage paths, and dense disk spindle counts. All SAN software features are also built in and include thin provisioning, snapshots, remote copy, and Network RAID for high availability. Being fused into the BladeSystem architecture also provides a density and footprint advantage further saving costs. The P4800 SAN for BladeSystem is designed for departmental or enterprise environments with greater than 500 users. Scaling from 2 nodes and 70 disks (see Figure 14) all the way to 8 nodes with 280 disks in a single BladeSystem c7000 enclosure with associated storage drawers or 16 nodes with 560 disks across multiple BladeSystem c7000 enclosures and storage drawers insures I/O performance is outstanding for the large scale VDI deployments that the P4800 G2 addresses. The ability to make private storage networks where storage traffic remains within the Virtual Connect domain means one administrator can manage the entire infrastructure stack without the need for specialized storage skill sets or in depth network administrator involvement. Figure 14: The HP P4800 G2 SAN Solutions for BladeSystem The P4800 G2 is a SAN and as such may serve as the boot device for all blades within the infrastructure stack. This eliminates local disks from the server and, with proper Virtual Connect domain setup, makes adds, moves and changes a breeze. Whether you are looking to place persistent VMs or non-persistent VMs, the P4800 offers up the mix of capacity with thin provisioning and performance required for VDI. 19

20 Benefits for VDI: The convergence of not only the storage but also the transport layer into a single, simple to manage infrastructure means intensive multi-team and multi-skill set involvement is a thing of the past. This leads to more rapid deployment and faster response to business driven change requests. The fact that the fabric is integrated across high speed electrical links leads to performance that is both impressive and scalable. Accelerated storage with HP P4000 Virtual SAN Appliance Software HP provides accelerated storage for VDI deployments utilizing the management servers that already exist within the enclosure combined with either HP IO Accelerator modules for HP BladeSystem or the HP SB40c storage blade with hot plug SSDs. Figure 15 shows the SB40c couple to a BL460c management server. Figure 15: HP P4000 Virtual SAN Appliance Software on ProLiant BL460c G7 and HP SB40c Storage blade With this approach, at least two management servers are virtualized and a virtual storage appliance is created. This appliance is given control of either SSDs or IO Accelerator modules. Data written to one appliance is replicated to a second appliance creating data availability while simultaneously providing scaling. This storage layer scales independently of all other storage layers and storage heads. Like the P4800 G2 SAN Solutions for HP BladeSystem, the accelerated blade utilizes the Virtual Connect domain communication paths to communicate with initiators resulting in a high speed, low latency storage network that requires no involvement from the SAN or network team when configured to not egress the domain. Benefits for VDI: Utilizing P4000 Virtual SAN Appliance Software (VSA) for selective use as a tiered storage layer in VDI implementations lowers cost and complexity and allows for right sizing the storage selection at other tiers in terms of cost and performance rather than locking into a single storage platform for all tiers. VSA is even managed from the same console as the other P4000 storage tiers and scales simply to meet demand. 20

21 Direct Attached Storage for HP BladeSystem The HP 600 Modular Disk System brings scalable direct attached storage (DAS) to blades, a critical piece to achieving the most cost effective VDI implementation. By architecting application and user virtualization into VDI implementations as HP suggests, the possibility of direct attached storage replacing pieces of the SAN is a reality. This brings with it the added benefit of isolating the potential storage bottleneck to a single server. With cost effective direct attached storage that is expandable on a per server basis, a bottleneck becomes unlikely. Figure 16: HP 600 Modular Disk System (MDS600) which provides direct attached storage for BladeSystem With DAS, the disposable pieces of the VDI implementation that are normally placed on SAN are moved to inexpensive, but high performance disks. This results in a per user cost that is a fraction of the cost of SAN solutions while delivering every ounce of performance those solutions provide. Even the servers can boot from the DAS infrastructure eliminating the disks from the local server itself. Benefits for VDI: While DAS is not an answer for every scenario, selecting it as the storage basis for the scenarios that do make sense reduces the overall implementation costs of VDI. With HP you have the added benefit of platform convergence which means whether you implement DAS or SAN, storage acceleration is always available and the management of the overall infrastructure doesn t change. When properly implemented, DAS also eliminates widespread storage bottlenecks and introduces cost effective storage scaling on a per server basis to VDI. 21

22 User data storage Data centers today rely on a variety of storage solutions to support different networks, applications and users. Flexibility and scalability are key to meeting these various and often changing requirements. HP X3800 G2 Network Storage Systems meet the requirements of any size data center. For example, the HP X3800 G2 Network Storage Gateway, a network-attached storage (NAS) appliance, with Windows Storage Server 2008 R2 and Emulex connectivity, provides large data centers the performance, flexibility and scalability they need to support tens of thousands of users. Figure 17 shows the X3800 G2. Figure 17: X3800 G2 Network Storage Gateway Benefits for VDI: The X3800 G2 integrates into existing Microsoft Windows management processes and when connected to expandable, scalable storage such as the HP P4500 G2 SAN creates a highly available repository for end user data. This brings user data into data center backup plans while keeping management simple. 22

23 Management VDI environments span multiple groups, virtualized resources and hardware stacks across the data center. HP Insight Control management software ships as one, installs as one, and provides a single, integrated management solution for both physical and virtual servers. Insight Control uses a central management console, HP Systems Insight Manager, providing a comprehensive and intuitive interface for running an entire set of server management tasks. HP Insight Control also offers the flexibility to integrate seamlessly into Microsoft System Center and VMware vcenter environments if your IT environment has already standardized on those management platforms. HP Insight Control delivers server lifecycle management functions tuned for HP ProLiant and HP BladeSystem environments in the areas of system health, performance monitoring, remote control, automated deployment and migration, power management, and virtualization management. These capabilities help improve IT operations by streamlining IT service rollouts, reducing incidence of unplanned downtime, increasing power efficiency, and delivering unified management of physical and virtual infrastructure. Benefits for VDI: To truly achieve single administrator abilities for VDI in the data center, you need to be able to control the infrastructure simply and respond to events rapidly. Insight Control brings the ability to deploy, monitor, control and optimize the infrastructure stack from anywhere the network is available. The ability to integrate with existing management platforms enhances the value. Access devices HP Thin Clients When you make the decision to move to Client Virtualization, you will face the inevitable question as to whether to keep your PC or to buy a thin client? Isn t that really the question? In most cases, you likely already own the PC, so why not repurpose its use for Client Virtualization? Many customers evaluate that same decision. Beyond just the normal life span answer, where thin clients are known to have a much longer lifecycle, you should know that PCs run hotter, run louder, require more electricity and are not purpose-built for Client Virtualization. Let s take a look at some simple statistics. 23

24 From the HP Carbon Footprint Calculator, you can see in Figure 18 some of the results, using as the cost per KWh and 1.36 lb COe/KWh (greenhouse gas emission factor), comparing the HP 6005 commercial desktop to an HP t5565 thin client. Figure 18. Estimated Electricity Costs and Carbon Footprint Electricity savings of just over $8,000 over the 3 year period (you may likely see much higher results with similar deployments, per the note below), carbon savings equivalent to 112,000 miles driven, and more than 50% of the pounds of CO2e. NOTE: The HP 6005 is newer than you are likely using today, so your results should be even greater. Thin clients also have availability benefits over a repurposed PC since there are no moving parts to fail. Nearly one of every four PCs run the risk of failing beyond year 3 of their expected life, increasing maintenance costs, compared to less than 3% for thin clients. This also speaks to reliability since PCs depend on fans and disk drives, increasing their risks of failure, making thin clients more reliable for productivity environments. Now, in leveraging one brand across your IT infrastructure, you can choose HP Thin Clients, the ideal client devices for Client Virtualization deployments providing increased security, simplified management, and strong performance. HP offers a full portfolio of Thin Clients with varying capabilities enabling IT to optimize both end-user experience and their IT budget by deploying a mix of client devices. 24

25 Figure 19. HP s Thin Client Portfolio The Simplest Solution at our Best Price HP t5335z HP ThinPro Essential Mainstream Enhanced Features for Mainstream Business Use HP t5500 Series HP ThinPro Windows CE 6.0 Windows ES 2009 Flexible Powerful, Flexible, Innovative! HP t574x Series HP ThinPro Windows ES 2009 Windows ES 7 Breakthrough size, performance & cost Full feature set thin client at a great value Desktop performance for PC replacement solutions Let s take a closer look at three of these HP Thin Clients: Zero Clients: HP t5335z (Essential) and t5565z (Mainstream) Smart Clients: Starting in 2011, HP will enhance its Thin Client portfolio with the new HP t5335z and t5565z Smart Clients. HP Smart Clients deliver an unmatched combination of flexibility and affordability to the zero client market -- unlike other zero clients available today, HP Smart Clients unique device purposing delivers a nocompromise IT experience with the choice to deploy for Citrix XenDesktop via HDX, or VMware View or Microsoft Session Virtualization via rdesktop with the option to migrate to any of the three software architectures at any time. Mainstream Line: HP t5565: A Mainstream series client, the t5565 is designed for most business productivity applications typically found with task and productivity workers. Easy setup. Simple management. Preinstalled with HP ThinPro operating system this all-inclusive access device is designed to provide a single console interface for dashboard access to all local resources making setup straightforward. The HP t5565 is ENERGY STAR qualified and EPEAT registered. Flexible Line: HP t5740e: For knowledge workers who require use of more advanced applications and media streaming, the t5740e is packed with the latest technology, delivering true PC performance, flexible expansion and reliability in Client Virtualization environments. Plus, it comes preinstalled with Microsoft Windows Embedded Standard 7, meaning the latest Microsoft OS features. HP s strategy to differentiate through offering flexible hardware platforms and modular software, rather than proprietary or single-purpose devices, translates to targeting the largest and fastestgrowing ISV environments and use-cases, building on our leadership position. 25

26 t5745 t5740e t5570 t5565 t5550 t5325 System Figure 20. The HP t5740e Thin Client with integrated wireless and dual monitors Before moving off of the HP Thin Client discussion entirely, let us briefly address protocols. Protocols are a very important part of the total solution. There are over 30 vendors in this space, though the big 3 are Microsoft, Citrix and VMware. Differences between them are based on design decisions often due to heritage of the vendor and the technology. When we talk about the big 3, Microsoft Remote Desktop Protocol (RDP), Citrix ICA and VMware PCoIP, these protocols offer similar experiences, so the selection of which is more about what the customer is trying to do. And, increasingly, this really is a matter of what server OS is selected. Based on that fact, you can use the data in Figure 21 to see what protocols, with latest version numbers, are supported by which products in the HP Thin Client portfolio. Figure 21. ISV Portfolio matrix with HP Thin Clients by Operating System WES7 WES2009 WinXPe Windows CE ThinPro Protocol Preinstalled Add-On (Softpaq) Preinstalled Add-On (Softpaq) Preinstalled Add-On (Softpaq) Preinstalled Add-On (Softpaq) Preinstalled Add-On (Softpaq) Citrix ICA VMware PCoIP 4.01 n/a Microsoft RDP n/a n/a Citrix ICA VMware PCoIP n/a n/a Microsoft RDP 6.0 Citrix ICA VMware PCoIP Microsoft RDP n/a n/a Citrix ICA VMware PCoIP Microsoft RDP 7.0 Citrix ICA VMware PCoIP Microsoft RDP 7.0 None Citrix ICA VMware PCoIP Microsoft RDP n/a n/a NOTE: This data is subject to change as HP aligns with ISV time to market and all new releases. 26

27 The performance of all protocols is impacted by many things including (1) thin client graphics decompression performance, (2) server (or host ) performance, specifically, the host s ability to compress and transmit what the virtual PC is presenting, and (3) network bandwidth and network latency. The more demanding the task (for example, watching HD video full screen on a 24 monitor), the more capable all of the above variables must be to deliver a good end user experience. Another important consideration relating to the protocols selection is whether multimedia redirection is supported, which depends on the architecture. For RDP, it is achieved only with HP enhancements. There are many HP competitors with platforms competing against HP Thin Clients deploying Microsoft Remote Desktop Services (RDS) that would not have similar multimedia redirection capabilities. Take a moment to review the grid below to make note of those protocol features that are not supported if a selected architecture is chosen. For example, if you want multi-monitor support and are evaluating VMware View 4.6 with RDP protocol, you will get the support, but not Aero Glass. Figure 22. HP protocols matrix Features by environment, by product and by OS * Microsoft Citrix XenApp 5 FP2 XenDesktop 4 VMware View 4.6 Protocol RDP 7 RFX * ICA 12.0 RDP 7 PCoIP Multimedia Redirection HP RDP Enhancements (Redirects WMV files) Supported Citrix HDX MediaStream HP RDP Enhancements Host-rendered Flash Redirection HP RDP Enhancements extended to support Flash (in Internet Explorer) Host-rendered Citrix HDX MediaStream for Flash HP RDP Enhancements being extended to support Flash (in Internet Explorer)* Host-rendered USB Redirection HP RDP Enhancements Supported Citrix HDX Plug-n-Play HP RDP Enhancements Supported True Multimonitor support Supported, but not Aero Glass Supported Citrix HDX Plug-n-Play ** Supported, but not Aero Glass Supported VoIP or Bidirectional Audio Bi-directional audio only Bi-directional audio only Citrix HDX RealTime Bi-directional audio only Bi-directional audio only *** * This represents the most common optimized environments, your back-end could offer different experiences ** While Citrix TrueMonitor supports up to 10 monitors, HP thin clients support up to 4 monitors *** Requires Teradici-directional audio driver from Please note, device redirection generally means the ability to access resource types from within a remote connection session. The resource types and the access questions you will need to address: 1. Flash storage devices Can the thin client connection allow a remote application or OS to read from and write to a locally attached device or to/from a network share device that is mapped to the thin client? 27

28 2. USB. For a USB printer Can the thin client connection redirect print output from a remote application to a locally attached printer on the thin client? For other USB ports (local ports) Can an application on the remote system see the local ports and access devices on those ports? This ability enables access to different print and storage devices, but it can also enable functions like ActiveSync synchronization of a locally attached personal digital assistant (PDA) with the serverbased /calendar application. HP Services No matter where you are on your Client Virtualization deployment, HP has the right services to assist you. Just beginning your Client Virtualization journey? HP Client Strategy Services deliver custom roadmaps for the journey to a fully optimized desktop environment. Offerings include: Transformation Experience: This workshop simulates a complete transformation of the desktop to attain stakeholder buy-in, build a desktop strategy supporting business and IT objectives, and gain traction for desktop transformational projects; and creates an initial roadmap for achieving the desired state Business Benefits Roadmap: Provides a comprehensive view of the benefits of the organization s strategic client initiatives, develops key elements of the business case, and creates a strategic roadmap Strategic Architecture: Provides enterprises with a clear blueprint and roadmap for effective desktop solution implementation and integration into their IT corporate architecture Planning to implement Client Virtualization, but want the shortest path to the best plan? Client Virtualization Assessment One of the keys to a successful Client Virtualization implementation lies in understanding the current load and using characteristics of your existing physical desktops. To do that, it is necessary to have an accurate understanding of application usage, as well as resource usage, including CPU, memory and disk input/output. HP Client Virtualization assessment provides comprehensive pre-virtualization analysis, modeling and planning support for your enterprise and data centers. Instead of relying on sizing and performance models developed in a virtualization vendor s lab, HP Client Virtualization assessment is designed to assist you with: Analysis of the Microsoft Windows client devices in your environment Capacity planning and resource modeling based on your needs Planning for desktop operating system images required to virtualize desktops Greater understanding of application usage Identification of applications that could pose challenges for your virtualization project The result is a series of reports and recommendations for a properly planned and sized virtualization project. 28

29 Starting a Windows 7 migration and investigating Client Virtualization? Often the two go hand in hand. A good strategy for the future of client computing should encompass both migration and virtualization. Some populations are very well suited for Client Virtualization, and the business benefits of virtual clients outweigh the benefits of migration to a traditional client. Consider the HP Client Strategy services to help you optimize your choices. And, in the course of Windows 7 migration, some of the thousands of applications in your portfolio may not migrate well to Windows 7, or may not migrate at all. Often putting them into a virtualized environment with one of the Client Virtualization models is the simplest solution. HP s Windows migration services begin by evaluating your portfolio of applications and working through those decisions with you. Want HP to design and build your solution, and turn it over to you? HP Client Virtualization Consulting Services HP offers you a broad portfolio of services built on best practices and encompassing every phase of the virtualization lifecycle with a focus on strategy and business value. HP partners with key software suppliers for solutions that are built to include your choice of technologies and more importantly, experience implementing all of the varieties of Client Virtualization. In addition to the Strategy and Assessment services, HP s lifecycle of services includes: Solution Design focusing on integrated solution and architecture design, application compatibility, and system management Proof of Concept and Pilot verifying design and integration, as well as performance and user acceptance for virtual client infrastructure and Client Virtualization Solution Implementation deploying your solution leveraging best practices and roadmap, plus proof of concept learnings Ongoing support giving you the level of support you need when you need it Want HP to design and manage your client infrastructure solution? In today s Client Virtualization market space, there are a number of Client Virtualization options that are available from a variety of vendors ranging from internally sourced and managed solutions to fully-outsourced and managed services. Choosing the optimal solution can be difficult and requires careful consideration to make sure the solution fits your long term business goals. HP has researched, tested and delivered many of these solutions to our clients across the full range of Client Virtualization scenarios. From this knowledge and experience, we have designed and created a flexible and scalable service offering that designs, delivers and maintains three Client Virtualization models that cover the majority of our clients needs. The models included in the HP Client Virtualization Services portfolio are designed to offer maximum flexibility based on the business needs of our clients and the computing needs and requirements of specific user segments within our clients organizations. These models can be deployed and managed within the client s data center or hosted by HP in one or more of our data centers around the globe. HP Client Virtualization Services includes three compute models to meet your needs: Presentation Virtualization offering centrally hosted delivery of applications to the end user s device where the application functions as it would if it were installed locally on a traditional PC. This model serves task workers well who only need secure access to a small number of applications. Desktop Virtualization using non-persistent virtual desktops where multiple end users access a pooled image of standard applications and user personalization and applications are streamed down to the delivered image for the look and feel of a personalized desktop. This model appeals to knowledge workers who have more diverse and customized computing needs than task workers. 29

30 They access a variety of personal productivity applications and have multiple applications open at a time. Desktop Virtualization using persistent virtual desktops that provide each end user with their own private image in a virtual machine with the look, feel and functionality of a traditional laptop or PC and the capability of the user to install non-standard applications without change control. Performance users who require access to compute-intensive applications, require the administrative capability to upload non-standard applications or have special graphic intensive requirements can benefit the most from this model. HP also offers Thin Client Management within the Client Virtualization Services portfolio for those customers moving to more efficient thin devices along with virtualized application presentation and virtual desktop access. This service features automation tools for asset management as well as remote support. HP can deliver each of these models individually wherever your employees are, and to whatever device they re using. Equally important, we are able to simultaneously deliver all three models from a single Converged Infrastructure. While the operating system and software function the same way for the user, your IT organization benefits from lower costs, stronger security, better performance, and increased flexibility and scalability. Constructing the Reference Architecture The following sections address at a high level how the Converged Infrastructure layers of the reference architecture are put together to make the best platform for VDI. These sections are designed to give a clearer understanding of the placement and function of each of the layers. For specific instructions on implementing VDI on the reference architectures, consult the partner reference architecture documentation at Enterprise Figure 23 disseminates the components involved in building the enterprise reference architecture Converged Infrastructure, with SAN as the primary storage layer. Note that a 4 node P4800 G2 SAN is shown. The P4800 G2 SAN Solutions for BladeSystem scale between two and eight nodes within a single HP BladeSystem c7000 enclosure. Figure 23: VDI enterprise reference architecture components 30

31 Each of the components serves a role within the overall architecture. Figure 24 highlights each role as an example configuration. Depending on the solution there are opportunities to further optimize this architecture with application delivery servers, Active Directory controllers and other systems that can benefit from incorporating their services into the tightly interconnected and high performance Virtual Connect domains. Figure 24: Functional view of the enterprise reference architecture with two c7000 enclosures and a four node P4800 G2 SAN Solution for BladeSystem 31

32 Including direct attached storage When architected to utilize non-persistent virtual machines, VDI becomes capable of a mixture of availability, cost and performance by utilizing direct attached storage. The advantage is the per user cost for storage drops while performance does not change. Important files and applications that the users access remain protected with the appropriate storage, but the disposable, temporary data they create at runtime is placed on DAS. Figure 25 disseminates the components involved in building the enterprise reference architecture with direct attached storage Converged Infrastructure. Figure 25: VDI enterprise reference architecture with direct attached storage components To save costs and improve power footprints each hypervisor host is booted from the same disks that host the disposable files. These disks are configured as a RAID10 set of between 2 and 8 disks depending on I/O load of the users and the overall number of users per blade. Sizing is planned on a per server basis. 32

33 Figure 26 highlights the various components involved in the solution. Note that shared, hot plug solid state disks are used in place of the I/O accelerators assumed in Figure 25. Either approach may be used to scale accelerated storage where required. Figure 26: Components within the DAS architecture Building hybrid models Ideally, a VDI implementation will optimize each layer for the best performance and price mix. This may include mixing direct attached storage and SAN. This model is recommended as it helps drive costs down while placing optimized hardware at each infrastructure layer. This section discusses the details of such an implementation. For specific information on implementing with your VDI software partner of choice, consult the HP partner reference architectures at Figure 27 shows a hybrid implementation. This configuration consists of two Virtual Connect domains. Each domain consists of four BladeSystem c7000 enclosures. The first Virtual Connect domain is connected to a six node P4800 G2 SAN Solution while the second utilizes direct attached storage. Each domain is managed by its own Insight Control server or virtual machine as well as a dual purpose three node accelerated storage and management server combination. This example configuration is capable of supporting up to 10,500 task workers 2 within three racks depending on your hypervisor and VDI software vendor. 2 HP s performance testing suggests up to 96 task worker VMs per ProLiant BL460c G7. Consult the End users and server performance section of this document for definitions of user types and expected sizings by server type. 33

34 34 Figure 27: A hybrid DAS and converged SAN VDI configuration housing up to 10,500 task workers

35 End users and server performance About end users As there is no universally agreed upon set of definitions around user types, HP has worked to place quantitative as well as qualitative measurements around user types and to help customers understand when a user of a particular type may end up in a different classification category. Task workers These workers are often defined as heavier users with other test methodologies. In essence, the profile for this user is fairly basic and includes the following characteristics: A minimal set of applications (the HP test bed used a mix of Microsoft Office, Microsoft Internet Explorer, Adobe Acrobat and a zip program). Only one application is open at a time. A minimal number of applications used during the course of a workday with generally 7 or fewer available. Interaction with the computer may be sporadic rather than constant. In addition, the following characteristics may be common: I/O counts of less than 10 at peak and 6 on average. HP has seen user counts in excess of 150 users on HP ProLiant BL460c G7 platforms when memory or other subsystems are not a factor in overall system contention. The type of users that enable these counts tend to be an ideal fit for Client Virtualization and may be adequately served by a variety of Client Virtualization technologies including VDI. It is common to see sizing numbers throughout the industry that are representative of HP s task worker workload. Examples. Call centers, users with primarily web based applications, manufacturing floors, healthcare workers and educators in environments with low levels of concurrency (many users on a host, but only a portion of those users active at a given point in time). Productivity workers HP defines a productivity worker as an office worker that exhibits all of the following characteristics: Multiple, non-web based applications open simultaneously. A range of more than 7 applications in use during the course of normal business with access to 12 or more. Frequent interaction with the computer. The core compute resource is a critical part of the end users job and they will interact with it on a regular basis. In addition, the following characteristics may be common: I/O counts of greater than 6 IOPs per user with counts as high as 16 IOPs. Compute services in the operating system image that generate CPU or I/O on a scheduled basis. Productivity users are a very common user type and are perhaps ideally suited for VDI as they tend to exhibit characteristics that are cost effective for VDI while not being easily addressed by other Client Virtualization approaches. Examples. Healthcare workers in environments with high concurrency, administrative professionals, business support staff, data entry workers. NOTE: HP expects that servers of similar configuration will perform in a similar fashion and thus we believe our numbers to be useful in comparing competitive systems of identical build. For customers who wish to be aggressive with their user count assumptions, HP suggests that a number between the task worker and productivity worker numbers shown in Table 2 in this document be used for early 35

36 planning stages prior to a proper pilot or assessment engagement. This recommendation is for two reasons. First, optimistic sizing without properly understanding user type may lead to a dramatic difference between planned cost and actual outlay. Second, application stacks may play just as large of a role as user behavior/type and this can easily shift workers from one category to the next. Both of these hurdles as well as issues around user selection and application delivery decisions can be addressed by HP s Client Virtualization Analysis and Modeling (CVAM) service. Knowledge workers A knowledge worker may best be categorized as a more powerful productivity worker. User counts in testing that HP conducts tend not to differ greatly from productivity workers. These users tend to exhibit the following characteristics: Multiple, non-web based applications open simultaneously. A large range of applications in use during the course of business. Frequent interaction with computer as a core part of their job. Manipulation of large data sets. In addition, the following characteristics may be common: I/O counts in excess of 15 IOPs per user on average with peaks reaching 25 or more IOPs seen as normal. There are applications in the operating system image that generate large amounts of CPU, memory or disk utilization. These users tend to be more expensive than task or productivity workers from a purchase and compute resource standpoint, but the data they work with, needs for security or the nature of their roles may justify placing them in a VDI environment. Examples. Finance (not traders), marketing professionals, engineers. Developers and other power users HP does not create nor has it found any references to VDI test beds that can reflect the activities of developers or other power users. This precludes any detailed discussion about lab findings for sizing. It has been common in the field to find developer communities supported at a rate of users on 8-12 core, 2 socket Intel Xeon based servers such as those mentioned in this document. HP highly recommends a services engagement to evaluate the fitness of developer communities for VDI. VDI desktops are an excellent method to host platforms for test and development requirements for developers who currently have multiple desktops provisioned. Power users may benefit substantially from HP blade workstations which deliver the benefits of data center data centralization and control while providing workstation capabilities to the end user. Configuration of recommended platforms HP has found that the platforms recommended in this document have proved to be exceedingly popular as well as functionally outstanding in VDI environments. Each platform is described by its as tested configuration as well as when it is best selected over other platforms. HP ProLiant BL460c G7 For this test, the ProLiant BL460c G7 is configured with 2 Intel Xeon X5660 processors running at 2.80 GHz with 12 x 8GB Dual Rank x4 PC DIMMs for a total of 96GB of memory. Drive configuration will vary based on storage architecture. Consult the reference architecture document related to your VDI vendor to determine the best drive configuration for your environment. The ProLiant BL460c scales best in environments where productivity or task workers may be served by virtual machines with 1 to 1.5 GB of memory or with larger memory configurations where customers embrace memory enhancement technologies at the hypervisor level. 36

37 HP ProLiant BL490c G7 For this test, the ProLiant BL490c G7 is configured with 2 Intel Xeon X5650 processors running at 2.66 GHz with 18 x 8GB Dual Rank x4 PC DIMMs for a total of 144GB of memory. Consult the reference architecture document related to your VDI vendor to determine the best drive configuration for your environment. The ProLiant BL490c G7 scales best in environments where task, productivity or knowledge workers may be served by VMs sized from 1.5 to 2GB of memory or with larger memory configurations where customers embrace memory enhancement technologies at the hypervisor level. HP ProLiant BL620c G7 For this test, the ProLiant BL620c G7 is configured with 2 Intel Xeon E processors running at 2.40 GHz and 24 x 8GB Dual Rank x4 PC DIMMs for a total of 192GB of memory. While this server is capable of much larger memory footprints with 8GB DIMMs, the overall per server user count with productivity and knowledge workers did not hit a memory capacity limitation. HP test methods and results HP utilizes a test methodology that seeks to evaluate the consistency of the end user experience. This consistency is tied to overall satisfaction by end users so HP finds it a critical measure of success. HP s methodology allows testing for more than one workload, protocol and application set. With the HP methodology, scripts are recorded individually using AutoHotKey with application launch times noted not just for when the process appears, but for when the user is first able to start interacting with the application. Initial application timings are measured based on a single user and are judged to be of acceptable performance and representative of launch times on individual VMs, local desktops and laptops. Timings are then increased by between 20% and 100% for specific monitored applications and written statically into scripts. Increases in application launch timing cause the perception of the solution as slow once the user has established a baseline of expected performance. Each script uses a variety of applications and in an array of combinations. The number of applications open, the frequency with which users interact with the applications and the speed with which data is fed to the applications are among numerous variables within the scripts. There are more than 80 scripts written with 8 being called task workers, 12 called knowledge workers and the remainder tied to productivity workers. A task worker script will have a minimal number of applications open and in use while a knowledge worker will have a larger number on both counts. To arrive at a final sizing, a predefined mix of these scripts is run and the respective hypervisor performance tools are used to record server side performance data. These tests are monitored during playback and server sizing is revealed by script failures at test time. In other words, when a script cannot complete as designed with other scripts running within the prescribed acceptable timeframe, the server has reached capacity. These failures show up in a variety of forms but are best described as unintended application behaviors. This equates to an end user s experience no longer being able to meet expectations. All scripts are run after login traffic has had a chance to settle to eliminate any effect. Logins, HA failures and boots are studied separately. Applications are set throughout the tests to launch within a critical range. These times are different than application response times to specific actions that are found in other test methodologies. Those ranges are listed in Table 1 below. This table also shows relative usage of media, number of concurrent applications per user and the total number of different scripts used in testing. 37

38 Table 1. Test characteristics by type User Type Number of Applications Available Critical Application Launch Times Concurrent Applications Media Usage Number of Different Scripts Task Worker seconds 1 Minimal 8 Productivity Worker seconds % 72 Knowledge Worker seconds % 80 Based on testing and field experience, HP observed the following user ranges as shown in Table 2. These counts assume Microsoft Windows 7 Professional 64-bit with a 37GB shared central image and enough memory and disk to satisfy requirements for the specified workload within the virtual machine. All applications were installed locally in the VM. See the section entitled Performance factors outside of user type in this doc to address the effects this has on sizing. For sizing purposes, a 1GB VM is assumed for task workers, 1.5GB for productivity workers and 2GB for knowledge workers. See the previous sections for individual hosts to identify their configurations at time of test. It should be noted that HP s productivity and knowledge worker workloads are very memory intensive and thus show diminishing returns as larger amounts of memory are placed in a server. NOTE: The test image of 37GB was sized to incorporate all test applications while leaving 10GB of free space for future expansion and patching of the master image. Customer image sizes will vary based on application delivery technologies, optimizations applied during build time and how profiles are utilized in the environment. Table 2. Per server user counts by host and type User Type HP ProLiant BL460c G7 HP ProLiant BL490c G7 HP ProLiant BL620c G7 Task Workers (1GB) Productivity Workers (1.5GB) Knowledge Workers (2GB) It is important to note that these results should not be compared to other results using different test methodologies. Based on experience and virtualization benchmark results, it is expected that these results will be valid within an acceptable error range for systems running the same versions of software and configured with the same processors and memory. No lab sizing can be completely matched to a given customer implementation so these and any other sizing results are for early planning purposes only. An engagement with HP Services is recommended prior to a full VDI implementation to present the clearest picture of your individual environment. 38

39 Table 3 presents performance in a fashion that may be useful when evaluating other solutions. This table highlights the number of servers (in parentheses) and maximum user types in a fully populated 42U rack. Table 3. Cumulative user counts by host and type for a fully populated 42U rack Task Workers (1GB or less) Productivity Workers (1.5GB) Knowledge Workers (2GB) HP ProLiant BL460c G7 (64) HP ProLiant BL490c G7 (64) HP ProLiant BL620c G7 (32) Per platform numbers versus users per core HP uses numbers based on a particular platform rather than users per core. The reason for this is that individual platforms produce very different user per core numbers largely based on memory architecture and core count. Scaling decreases as you move to larger socket systems and thus user per core counts will decrease. Couple this with the fact that core counts range from 4 to 12 on systems with different memory architectures and it is possible to set unrealistic expectations. Performance factors outside of user type While the recommendations in this section make good generic planning numbers based on defined user types, there are other factors that may increase or decrease user counts on a host. These factors may be considered to help refine pre-implementation planning estimates. Use case Consider a shared access model for VDI. A simple set of applications is assumed to be installed on stateless VMs. In a shared access model, many more users exist than clients available to service them. As a result, a continuous queue forms where every time a user logs out, a new one will log into the client. The VM itself is generally reset each time. In a scenario where usage is short at each logon, this creates a continuous login storm of sorts that can drive I/O to extremely high levels. In addition to excessive storage utilization it also consumes CPU resources on host systems and may result in lower overall performance despite the fact that users are performing simple tasks while logged on. This is a single example of a use case that may have unintended consequences. Understanding how end users will interact with your VDI implementation may better help with estimating user types. Image The base image is a common area for unexpected sizing issues. An image that is loaded with scheduled services, extraneous programs and is generally un-optimized to be shared among a large number of users is a problem before the first VM is deployed. Such images reduce user counts on a per server and per storage target basis and can raise VDI costs considerably. As an example, consider a virus scan scheduled in a fully provisioned VM. For extremely dense environments with large numbers of users, scheduling virus scans inevitably means that multiple scans will need to occur at any given point in time. The net effect of having this service built into the image is an almost continuous drop in quality of service as shared resources are strained from the scan. Each VDI solution vendor has recommendations for tuning system images that should be followed. 39

40 Application set The choice of what applications to host in a VDI environment as well as how to present them to the VM can have a considerable effect on the overall performance of the VM as well as the characterization of the end users. In some circumstances, users that require a particularly difficult application may simply need to be removed from consideration as viable for a VDI environment. In other cases, it may mean that the applications need only be removed from the local VM. Application virtualization technologies may allow these apps to be streamed from other systems where the I/O and execution is offloaded. This may, in effect, raise user counts on the hypervisor host and allow task worker VM counts with productivity workers. Virtualization or disaggregation of applications from the VM is recommended for performance and management/control reasons. Storage contention In some environments, storage can become a point of contention. This may manifest itself with higher than expected user I/O or vastly different patterning than expected. The more centralized the storage infrastructure, the greater the chance that I/O will become a limiting factor in VDI implementations. A balance of distribution of load while minimizing targets is desirable. User behaviors Many IT shops are unaware of user patterns and may not have an in depth understanding of exactly what applications are in use and by whom. Making assumptions about user type may in some circumstances be justified. An example is the task worker described in this section. If there is truly only a limited application set and workers do not require a great deal of local storage then it is safe to call them task workers. Some job categories may be seen as task workers when in fact the role is far more complex than imagined. IT administrators need to be aware of work patterns to insure proper sizing. Under all of these circumstances HP can help with your move to VDI. See the section in this document about HP s Client Virtualization Analysis and Modeling (CVAM) service. Memory optimization technologies Hypervisor vendors have brought memory optimization technologies to market that help counteract the limitations of physical memory capacity. HP s test methods for VDI tend to exercise memory in a fashion that promotes excessive memory latencies (and thus decreased user response times) or, as is the case with task workers, exhausts CPU resources prior to using all physical memory available on the host. Memory behaviors will vary substantially based on workload and with proper image planning and design it is expected that you can benefit from memory optimization via increased user densities per host. HP Client Virtualization Analysis and Modeling service The most accurate way to get a clear understanding of user types and extraneous factors effecting a VDI implementation is to use data from your environment. The HP Client Virtualization Modeling and Analysis service is a five-week engagement that will help you understand the fitness of your environment and users for a move to VDI as well as help you understand the peaks and valleys they will produce on your infrastructure. At the end of the service, HP will have helped eliminate the guesswork and will provide you with information you need to move forward. 40

41 For more information For a central repository of all HP Client Virtualization documents, go to HP Client Virtualization Analysis and Modeling Service, HP Insight Control management software, Call to action Before adopting Client Virtualization, dig deep. This document is one in a series of documents HP is producing to help address questions around Client Virtualization and VDI in general. All of the documents are published at: To help us improve our documents, please provide feedback at Copyright 2011 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice. The only warranties for HP 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. HP shall not be liable for technical or editorial errors or omissions contained herein. Microsoft and Windows are U.S. registered trademarks of Microsoft Corporation. Intel and Xeon are trademarks of Intel Corporation in the U.S. and other countries. ENERGY STAR is a registered mark owned by the U.S. government. Adobe and Acrobat are trademarks of Adobe Systems Incorporated. 4AA3-4348ENW, Created May 2011, Updated June 2011, Rev. 1