HPE Nimble Storage ReferenceArchitecture for Citrix XenDesktop

ContentsOverview.5Audience.5Solution Summary.5HPE Nimble Storage Technical Overview.6All-Flash Array Hardware.6Universal Hardware Architecture.7Predictive Flash Platform.7HPE NimbleOS Architecture.8Unified Flash Fabric.8Transparent Application Migration.9HPE InfoSight.10VMVision for Hypervisor and VMware Monitoring.11Solution Design.12HPE Nimble Storage Design Fundamentals.12Citrix XenDesktop Design Fundamentals.12Citrix XenApp 7.11.12Citrix XenDesktop 7.11.13Citrix Provisioning Services 7.11.14Citrix Desktop Studio for XenApp 7.11.15Benefits for Desktop Administrators.15Citrix Provisioning Services Solution.16Citrix Provisioning Services Infrastructure.16High-Level Storage Architecture Design.17Solution Hardware and Software.18Solution Components.18Logical Architecture.18VLANs.20VSANs.20VMware Clusters.20Data Storage Layout.21Testing Methodology and Success Criteria.22Testing Procedure.22Copyright 2018 by Hewlett Packard Enterprise Development LP. All rights reserved.

Pretest Setup.22Test Run Protocol.22Success Criteria.23VSImax.24Server-Side Response Time Measurements.24Calculating VSImax v4.1.x.24Test Results.28Superior Performance.28Storage Test Case I: 5000-User Testing with PVS RAM Cache Enabled.28Storage Test Case II: 5000-User Testing with PVS RAM Cache Disabled.29Storage Test Case III: 2500 XenDesktop Session Boot Storm.30Absolute Resiliency and Nonstop Availability.31Storage Test Case IV: HPE Nimble Storage Controller Failover with an Active Workload.31Storage Test Case V: Multiple SSD Failures with an Active Workload.34Moving Volumes Across Arrays with HPE Nimble Storage Transparent Application Migration.38How to Use Transparent Application Migration.38HPE Nimble Storage Predictive Analytics and Monitoring.40HPE InfoSight Predictive Analytics.40Overview Tab.40Performance Tab.42HPE InfoSight VMVision.43How to Configure VMVision Monitoring.43Types of VMVision Monitoring Information.45About the Author.48Bharath Ram.48Version History.49Copyright 2018 by Hewlett Packard Enterprise Development LP. All rights reserved.

Documentation Feedback Copyright 2018 Hewlett Packard Enterprise Development LP. All rights reserved worldwide.NoticesThe information contained herein is subject to change without notice. The only warranties for Hewlett PackardEnterprise products and services are set forth in the express warranty statements accompanying such productsand services. Nothing herein should be construed as constituting an additional warranty. Hewlett Packard Enterpriseshall not be liable for technical or editorial errors or omissions contained herein.Confidential computer software. Valid license from Hewlett Packard Enterprise required for possession, use, orcopying. Consistent with FAR 12.211 and 12.212, Commercial Computer Software, Computer Software Documentation, and Technical Data for Commercial Items are licensed to the U.S. Government under vendor's standardcommercial license.Links to third-party websites take you outside the Hewlett Packard Enterprise website. Hewlett Packard Enterprisehas no control over and is not responsible for information outside the Hewlett Packard Enterprise website.AcknowledgmentsIntel , Itanium , Pentium , Intel Inside , and the Intel Inside logo are trademarks of Intel Corporation in theUnited States and other countries.Microsoft and Windows are either registered trademarks or trademarks of Microsoft Corporation in the UnitedStates and/or other countries.Adobe and Acrobat are trademarks of Adobe Systems Incorporated. Java and Oracle are registered trademarks of Oracle and/or its affiliates.UNIX is a registered trademark of The Open Group.Publication DateFriday January 26, 2018 11:09:04Document IDkqg1501192545972SupportAll documentation and knowledge base articles are available on the HPE Nimble Storage Support site at To register on HPE InfoSight, click the Enroll Now link on the main page.Email: [email protected] all other general support contact information, go to pyright 2018 by Hewlett Packard Enterprise Development LP. All rights reserved.

Documentation FeedbackOverviewOverviewThe virtual desktop infrastructure (VDI) solution described in this guide is a predesigned, best-practice datacenter reference architecture that is built on HPE Nimble Storage all-flash arrays. The architecture wasvalidated for an up-to-5000-seat mixed workload that consisted of the following types of virtual desktop: Citrix XenDesktop persistent virtual desktopsXenDesktop nonpersistent pooled virtual desktopsCitrix XenApp server-based virtual desktop sessionsThe 100% virtualized solution uses VMware vSphere 6.0, Update 2, on an HPE Nimble Storage AF5000array with a Fibre Channel (FC) SAN architecture. The nonpersistent virtual desktops are powered by CitrixProvisioning Services (PVS) 7.11 with a mix of desktop software: XenDesktop 7.11, which supports both persistent and nonpersistent Windows 10 virtual desktopsXenApp 7.11 hosted, shared-server-based Windows Server 2012 R2 desktops, which provide unparalleledscaling and management simplicityThe solution was tested with the following desktops provisioned on an HPE Nimble Storage AF5000 array: 1000 XenDesktop Windows 10 persistent desktops2500 XenDesktop Windows 10 nonpersistent pooled desktops1500 XenApp Windows Server 2012 R2 RDS server-based desktop sessionsThe tested architecture provides outstanding virtual desktop end-user experience as measured by the LoginVSI 4.1 Knowledge Worker workload running in benchmark mode. The 5000-seat architecture functions asa large-scale building block that can be replicated to confidently scale out to tens of thousands of users.Where applicable, this guide presents best practice recommendations and sizing guidelines for deployingthe solution.AudienceHPE customers and partners can use the information in the guide to build an infrastructure that delivers ITefficiency and enables IT innovation. The audience for this document includes, but is not limited to, salesengineers, field consultants, professional services staff, IT managers, partner engineering, and customers.Solution SummaryThis document describes a complete virtual desktop architecture that uses the HPE Nimble Storage platformfor Citrix XenDesktop. The described implemenation provides test cases and reference architectures forsystem scalability, self-contained functioning environments, and multiple workloads on HPE Nimble Storagehardware. This guide focuses on the validation of the following test cases and use cases: Full-scale 5000-user mixed workload scalability for Citrix XenDesktop 7.11 pooled virtual desktopinfrastructure (VDI) (50%), persistent VDI (20%), and remote desktop session host (RDSH) (30%)Full-scale, mixed-workload scalability over a soak test period of 24 hours (minimum) during which LoginVSI was run in a steady-state manner for an extended duration5000-user mixed-user workload performance testing that showcased HPE Nimble Storage metrics suchas bandwidth, IOPS, and submillisecond latency during peak workloadHPE Nimble Storage absolute resiliency and nonstop availability through controller failover and drivefailure tests during an active workloadHPE Nimble Storage features such as transparent application migration, HPE InfoSight, and VMVisionCopyright 2018 by Hewlett Packard Enterprise Development LP. All rights reserved.5

HPE Nimble Storage Technical OverviewDocumentation FeedbackHPE Nimble Storage Technical OverviewAll-Flash Array HardwareThe HPE Nimble Storage AF-Series array family starts at the entry level with the AF1000 model and expandsthrough the AF3000, AF5000, and AF7000 models up to the AF9000 model at the high end. All-flash arrayscan be upgraded nondisruptively from the entry level all the way to the high-end array model.For more information, see HPE Nimble Storage All Flash Arrays.Figure 1: The HPE Nimble Storage all-flash array familyFigure 2: All-flash array front viewCopyright 2018 by Hewlett Packard Enterprise Development LP. All rights reserved.6

Universal Hardware ArchitectureDocumentation FeedbackFigure 3: All-flash array rear viewUniversal Hardware ArchitectureHPE Nimble Storage arrays are built on a universal hardware platform with modular components. Allcomponents, including controllers, can easily be upgraded nondisruptively by the customer or an HPErepresentative. The universal hardware architecture spans both all-flash and adaptive flash arrays, givingHPE customers maximum flexibility and reuse of their array hardware.Figure 4: HPE Nimble Storage universal hardware architecturePredictive Flash PlatformThe HPE Nimble Storage predictive flash platform enables enterprise IT organizations to implement a singlearchitectural approach that dynamically caters to the needs of varying workloads. Driven by the power ofpredictive analytics, predictive flash is the only storage platform that optimizes across performance, capacity,data protection, and reliability within a dramatically smaller footprint.Copyright 2018 by Hewlett Packard Enterprise Development LP. All rights reserved.7

HPE NimbleOS ArchitectureDocumentation FeedbackPredictive flash is built on the HPE Nimble Storage Cache Accelerated Sequential Layout (CASL) architecture,the HPE NimbleOS operating system, and HPE InfoSight, the HPE cloud-connected predictive analytics andmanagement system. CASL and HPE InfoSight form the foundation of the predictive flash platform, whichenables the dynamic and intelligent deployment of storage resources to meet the growing demands ofbusiness-critical applications.HPE NimbleOS ArchitectureThe HPE NimbleOS operating system is based on the patented HPE Nimble Storage CASL architecture.CASL leverages the unique properties of flash and disk to deliver high performance and capacity—all withina remarkably small footprint. CASL scales performance and capacity seamlessly and independently.Unified Flash FabricThe HPE Nimble Storage unified flash fabric is a single consolidated architecture that enables flash for allenterprise applications. In the past, enterprises have been forced to choose between hybrid flash and all-flasharrays. That is no longer the case with the HPE Nimble Storage unified flash fabric. The unified flash fabricenables flash for all enterprise applications by unifying HPE Nimble Storage all-flash and adaptive flash arraysinto a single consolidated architecture with common data services.Figure 5: HPE Nimble Storage unified flash fabricThe unified flash fabric provides many significant benefits: Thin provisioning and efficient capacity utilization. With CASL, capacity is consumed only as data iswritten. CASL efficiently reclaims free space on an ongoing basis, preserving write performance withhigher levels of capacity utilization. This strategy avoids the fragmentation issues that hamper otherarchitectures.Accelerated write performance. After writes are placed in NVDIMM (made persistent and mirrored tothe passive partner controller), they are acknowledged back to the host and sent to solid-state disk (SSD)at a later time (generally when there is a full stripe to be written). As a result, writes to an HPE NimbleStorage array are acknowledged at memory speeds.Maximized flash write cycles. By sequencing random write data, HPE NimbleOS sends full stripes ofdata to SSD. Compressing and deduplicating the data inline minimizes the data footprint on the disk. Inaddition, because the data sent to disk is of variable block size, HPE NimbleOS does not have to breakit into smaller, fixed-sized chunks to be placed on SSD. Therefore, it is sent to SSD efficiently. This efficiencyenables HPE Nimble Storage arrays to maximize the deployable life of a flash drive by minimizing writewear on the flash cells.Read performance. Because all reads come from SSD, HPE NimbleOS and HPE Nimble Storage all-flasharrays deliver submillisecond read latency and high throughput across a wide variety of demandingenterprise applications.All-flash arrays. HPE Nimble Storage all-flash arrays use only SSDs to store data. As a result, all readoperations come directly from the SSDs themselves, delivering extremely fast read operations. All writesare also sent to SSD, but because of the HPE NimbleOS architecture and the use of NVDIMMs to storeCopyright 2018 by Hewlett Packard Enterprise Development LP. All rights reserved.8

Transparent Application MigrationDocumentation Feedbackand organize write operations, all writes are acknowledged at memory speeds (just as with the HPE NimbleStorage adaptive flash arrays).HPE Nimble Storage all-flash arrays use triple-level cell (TLC) SSDs, which enable maximum flash storagedensity. Traditional SSD problems with write wear, write amplification, and so on are not an issue for thevariable block HPE NimbleOS architecture, which minimizes write amplification through its intelligent datalayout and management in the file system. Efficient, fully integrated data protection. All-inclusive snapshot-based data protection is built into theadaptive flash platform. Snapshots and production data reside on the same array, eliminating the inherentinefficiencies of running primary and backup storage silos. With its intuitive dashboards and proactivenotifications about potential issues, HPE InfoSight ensures that customers’ data protection strategies workas expected.Inline compression. CASL uses fast, inline compression for variable-application block sizes to decreasethe footprint of inbound write data by as much as 75%. When enough variable-sized blocks are accumulatedto form a full write stripe, CASL writes the data to disk. If the data being written is active, it is also copiedto SSD cache for faster reads. Written data is protected by triple-parity RAID.Inline deduplication. HPE Nimble Storage all-flash arrays include inline data deduplication in addition toinline compression. The combination of inline deduplication and inline compression delivers a comprehensivedata reduction capability that allows HPE NimbleOS to minimize the data footprint on SSD, maximizeusable space, and greatly minimize write amplification.Thin, redirect-on-write snapshots with SmartSnap. HPE Nimble Storage snapshots are point-in-timecopies that capture just-changed data, making it easy to store three months of frequent snapshots on asingle array. Data can be restored instantly because snapshots reside on the same array as the primarydata.Efficient replication with SmartReplicate. SmartReplicate sends only compressed, changed data blocksover the network for simple and WAN-efficient disaster recovery.Zero-copy clones. HPE Nimble Storage snapshots enable the quick creation of fully functioning copies,or clones, of volumes. Instant clones deliver the same performance and functionality as the source volume,which is an advantage for virtualization, VDI, and test or development workloads.Application-consistent snapshots. The HPE Nimble Storage VDI solution uses the Volume ShadowCopy Service (VSS) framework and VMware integration to create instant application-consistent andVM-consistent backups, using application templates with pretuned storage parameters.Flexible data encryption with SmartSecure. HPE NimbleOS enables individual volume-level encryptionwith little or no performance impact. Encrypted volumes can be replicated to another HPE Nimble Storagetarget, and data can be securely shredded at the volume level of granularity.Transparent Application MigrationEnterprises can deploy HPE Nimble Storage all-flash arrays, adaptive flash arrays, or a combination of bothto meet the varying needs of all applications. Because both types of arrays run the same HPE NimbleOS,management and functionality are identical, and arrays can be clustered and managed as one.The transparent application migration feature enables the administrator to transparently migrate volumes thatwere set up on an adaptive flash array to an all-flash array or vice-versa. If, for example, infrastructure volumesare housed on an adaptive flash array, and they need to be migrated to an all-flash array, moving themthrough transparent application migration can reduce the time required to rebuild the entire infrastructure.VDI infrastructure volumes can be migrated across arrays without affecting the consistency of the infrastructure.Copyright 2018 by Hewlett Packard Enterprise Development LP. All rights reserved.9

HPE InfoSightDocumentation FeedbackFigure 6: Transparent application migrationHPE InfoSightHPE InfoSight leverages the power of deep data analytics to offer customers precise guidance on the optimalapproach to scaling flash, CPU, and capacity for changing application needs while maintaining peak storagehealth. Using systems modeling, predictive algorithms, and statistical analysis, HPE InfoSight solves storageadministrators’ most difficult problems, ensuring that storage resources are dynamically and intelligentlydeployed to satisfy the changing needs of business-critical applications.At the heart of HPE InfoSight is a powerful engine that applies deep data analytics to telemetry data that isgathered from HPE Nimble Storage arrays deployed across the globe. More than 30 million sensor valuesare collected per day for each HPE Nimble Storage array. The HPE InfoSight engine transforms the millionsof gathered data points into actionable information that helps customers realize significant operational efficiencyin many ways: Maintaining optimal storage performanceProjecting storage capacity needsProactively monitoring storage health and receiving granular alertsProactively diagnosing and automatically resolving complex issues, which frees up IT resources forvalue-creating projectsEnsuring a reliable data protection strategy with detailed alerts and monitoringExpertly guiding storage resource planning, by determining the optimal approach to scaling cache andIOPS to meet changing SLAsIdentifying latency and performance bottlenecks through the entire virtualization stackDelivering transformed support experience from level-3 supportFor more information, see HPE InfoSight Predictive Analytics.Copyright 2018 by Hewlett Packard Enterprise Development LP. All rights reserved.10

VMVision for Hypervisor and VMware MonitoringDocumentation FeedbackVMVision for Hypervisor and VMware MonitoringVMVision provides a granular view of the resources that are used by every VM connected to an HPE NimbleStorage array. This feature makes it possible to correlate the performance of VMs in a datastore with insightsinto hypervisor and host resource constraints such as vCPU, memory, and network.VMVision helps in determining VM latency factors, whether from the storage, the host, or the network. It alsohelps in taking corrective action on noisy-neighbor VMs and reclaiming space from underused VMs. Everyhour, through the heartbeat mechanism, the correlated statistics are sent to HPE InfoSight for processing.No additional host-side agents, tools, or licenses are necessary for the feature to work.Figure 7: HPE Nimble Storage VMVisionCopyright 2018 by Hewlett Packard Enterprise Development LP. All rights reserved.11

Solution DesignDocumentation FeedbackSolution DesignHPE Nimble Storage Design FundamentalsThe solution design uses the HPE Nimble Storage all-flash array to provide block storage. A base configurationwith 46 TB of raw capacity was deployed in the array that was used for validation. HPE Nimble Storageall-flash arrays support the addition of up to two expansion shelves for a total raw flash capacity of 184 TB,with an effective capacity of approximately 680 TB (applying a 5:1 data reduction).This design uses 16 Gb fabric connectivity with two FC interface cards to provide 64 Gb of FC bandwidth percontroller. For additional FC bandwidth, a third FC card can be deployed on each controller; however, thisinterface is typically used for 10 GbE connections to other arrays in a scale-out cluster for data replicationtraffic. HPE Nimble Storage arrays support nondisruptive upgrades for increasing capacity (scaling deep),upgrading controllers (scaling up), or adding arrays (scaling out).The design uses an FC SAN to boot the servers. The service profile that is used to configure and deploy thecompute servers includes a boot policy that points to the HPE Nimble Storage array. The boot policy specifiesa primary and a secondary SAN path to the two controllers on the array where the boot volumes reside. Asecond boot policy is also configured, but with the primary and secondary paths reversed from that of thefirst boot profile. The second boot policy is used to load-balance the SAN boot across different paths whenmultiple servers are booting. This optional aspect of the design can be helpful in larger deployments fordistributing the load when multiple servers must be booted simultaneously.Each server has a dedicated 10 GB boot volume on the HPE Nimble Storage array. During the initial SANboot, the server attaches to all primary and secondary connections in both the active and the standby controller.This arrangement enables normal boot operations even when a controller or primary path is offline. The hostsare configured with the HPE Nimble Connection Manager and the path selection policy to optimize MPIOsettings and enable proper FC path management and failover connectivity.Citrix XenDesktop Design FundamentalsPrime reasons for moving to a virtual desktop solution include an ever-growing and diverse base of userdevices, complexity in the management of traditional desktops, the ongoing need for security, and even thegrowth in programs that let you bring your own device (BYOD) to work.Citrix XenDesktop 7.11 integrates hosted shared and VDI desktop virtualization technologies into a unifiedarchitecture that enables a scalable, simple, efficient, and manageable solution for delivering Windowsapplications and desktops as a service.Users can select applications from an easy-to-use “store” that is accessible from tablets, smartphones, PCs,Macs, and thin clients. XenDesktop delivers a native touch-optimized experience with HDX high-definitionperformance, even over mobile networks.Citrix XenApp 7.11The Citrix XenApp and XenDesktop application and desktop virtualization solutions are built on a unifiedarchitecture, so they are simple to manage and flexible enough to meet the needs of all users in an organization.XenApp and XenDesktop have a common set of management tools that simplify and automate IT tasks. Thesame architecture and management tools that are used for on-premises deployments are also used formanaging public, private, and flash cloud deployments.Citrix XenApp delivers multiple benefits: XenApp published applications, also known as server-based hosted applications. These applicationsare hosted from Windows servers to any type of device, including Windows PCs, Macs, smartphones,and tablets. Some XenApp editions include technologies that further optimize the experience of usingCopyright 2018 by Hewlett Packard Enterprise Development LP. All rights reserved.12

Citrix XenDesktop 7.11 Documentation FeedbackWindows applications on a mobile device by automatically translating native mobile-device display,navigation, and controls to Windows applications; enhancing performance over mobile networks; andenabling developers to optimize custom Windows applications for any mobile environment.XenApp published desktops, also known as server-hosted desktops. These inexpensive, locked-downWindows virtual desktops are hosted from Windows Server operating systems. They are well suited tousers such as call center employees who perform a standard set of tasks.VM-hosted applcations. These applications are hosted from machines running Windows desktop operatingsystems for applications that cannot be hosted in a server environment.Windows applications delivered with Microsoft App-V. These applications use the same managementtools that are used for the rest of a XenApp deployment.Citrix XenDesktop 7.11Citrix XenDesktop 7.11 includes significant enhancements to help customers deliver Windows apps anddesktops as mobile services while addressing management complexity and associated costs. The 7.11release includes the following enhancements: Unified product architecture for XenApp and XenDesktop—the FlexCast Management Architecture(FMA). A single set of administrative interfaces delivers both hosted-shared applications (Remote DesktopServices [RDS]) and co

The HPE Nimble Storage unified flash fabric is a single consolidated architecture that enables flash for all enterprise applications. In the past, enterprises have been forced to choose between hybrid flash and all-flash arrays. That is no longer the case with the HPE Nimble Storage unified flash fabric. The unified flash fabric