Reference Architecture - Citrix On Nutanix AHV

Citrix on Nutanix AHV
By Brian Martynowicz
Disclosure and Warranty

The information, concepts, and ideas contained in this document are the property
of Login VSI. Without limiting the rights under copyright, no part of this document
may be reproduced, stored in or introduced into a retrieval system, or transmitted
in any form or by any means (electronic, mechanical, photocopying, recording, or
otherwise), or for any purpose, without the express written permission of Login
VSI. Any product descriptions or representations in this document are for
identification purposes only and are not to be construed as a warranty of specific
properties or guarantee or warranty of any other type.
Login VSI shall assume no liability, either explicit or implied, for the
documentation. Information in this document, including URL and other Internet
Web site references, is subject to change without notice.
All sample code described in this document is provided by Login VSI for illustrative
purposes only. These examples have not been thoroughly tested under all
conditions. Login VSI, therefore, cannot guarantee or imply reliability,
serviceability, or functionality of these programs or code examples. All brand
names and product names used in this document are trademarks of their
respective holders and are recognized as such.
© 2017 Login VSI. All rights reserved.
Validated by Login VSI 1

Revision History
Revision Change Description Updated By Date
1.0 Document Creation Brian Martynowicz May 16, 2017
About the Author

Brian is a Senior Technical Consultant at Login VSI. Using 15 years of IT experience he provides technical
guidance on Login VSI and Login PI to customers and partners around the world. Over the last year and a half
working for Login VSI he has worked with enterprise customers, and vendors to demonstrate the value of testing
at all stages of VDI / SBC deployment. Brian is also a LinkedIn publisher since 2014, and contributes to a popular
Citrix blog called Citrixology. A native Baltimorean he enjoys living in the city, writing, playing/yelling at
competitive video games, working on cars, reading up on the latest technologies, and listening to rap music.

Contents
Executive Summary ....................................................................................................................................................4
Introduction ................................................................................................................................................................5
Nutanix Enterprise Cloud Platform Overview ............................................................................................................7
Solution Design ........................................................................................................................................................ 12
Citrix and Windows Optimization Best Practices .................................................................................................... 17
Solution Application: XenDesktop ........................................................................................................................... 22
Solution Application: XenApp .................................................................................................................................. 24
Results: Citrix XenDesktop on Nutanix .................................................................................................................... 29
Results: Citrix XenApp on Nutanix ........................................................................................................................... 33
Conclusion ............................................................................................................................................................... 35
Appendix .................................................................................................................................................................. 36
About Nutanix ............................................................................................................. Error! Bookmark not defined.

Executive Summary
In this reference architecture, Login VSI provides recommendations for designing, optimizing, and scaling Citrix
XenDesktop and XenApp deployments on Nutanix AHV. Citrix XenDesktop on Nutanix is a powerful solution that
offers ideal user experience, simple administration, and web-scale flexibility and economics. Citrix XenApp
transforms desktops and applications into a secure, on-demand service available to any user, anywhere, on any
device.
Login VSI provides the vendor independent evaluation of the Nutanix Enterprise Cloud platform, which contains
detailed performance and configuration information about the cluster’s ability to scale when used for XenDesktop
and XenApp deployments.
Citrix XenApp on Nutanix AHV Summary

The solution and testing data in this document derives from Citrix XenApp deployed on Nutanix AHV, running on
the Nutanix Enterprise Cloud Platform using MCS. Login VSI utilizes realistic workloads to help simulate the
conditions of a XenApp environment with MCS on Nutanix. Our testing demonstrates that host CPU resources
drive desktop user density, rather than concerns about I/O or resource bottlenecks—you can achieve densities of
over 250 Knowledge Worker desktops per node, with four nodes per 2RU appliance. We determined the pod sizes
by taking into consideration both performance and the additional resources needed to maintain N+1 failover
capabilities.
Citrix XenDesktop on Nutanix AHV Summary

The Nutanix Enterprise Cloud Platform offers high IOPS and low latency. Our testing demonstrates that host CPU
resources drive desktop user density, rather than concerns about I/O or resource bottlenecks—you can easily
deploy 150 knowledge worker desktops per node, with four nodes per 2RU appliance. In addition to desktop and
application performance reliability, when deploying XenDesktop on Nutanix you get unlimited scalability, data
locality, AHV clones and a single datastore. Nutanix also takes the Citrix commitment to simplicity to another level,
with streamlined management, reduced rollout time and lower operating expenses.

Introduction
Login VSI and EUC Environment Validation
Login VSI is the industry-standard performance validation tool that simplifies load testing, benchmarking and
capacity planning. It emulates real-world conditions using virtual user logons with customized workloads to ensure
the optimal end-user experience. Login VSI is agentless and has minimal infrastructure requirements, it works with
any mainstream broker, Windows operating system and Microsoft Office version.
There is no doubt that the digital workspace is transforming faster than ever due to the rapid innovation within
the digital work world. Compute technology alone has enabled a swift move into server virtualization, and with
the latest progress in flash storage technology we are seeing VDI and SBC adoption on the rise. One challenging
side-effect of this progress is that our virtualization administrators now must keep up with a pace of innovation
never seen before. For example, we are seeing the Windows 10 desktop OS updated at a pace that has caused
Microsoft to label it “Windows as a Service”. It sounds great, but what about those poor administrators who now
must update their base desktop OS twice a year? Updating an OS is straightforward… if you don’t have lines of
business applications that must also work without interrupting the productivity of the business itself—with the
added demand of a premium user experience.
If simply keeping up with business growth, keeping up with the innovation pace and making sure everything
continues to work despite all that change is a challenge—then when will you have time to focus on making sure
your end-users are happy and productive? This is where the flexibility, simplicity and speed from hyper-converged
architectures, like Nutanix, can make a massive difference.
If business growth is a factor, then an architecture that scales simply in a few clicks without the need to spend a
bunch of time managing the storage, compute and networking layer is paramount. From what we have seen, the
Nutanix Prism management framework delivers just that, and with integration to popular desktop provisioning
systems like Citrix MCS, administrators can save a lot of time and effort. This lets administrators to spend more
time testing Windows 10 updates, and performance tuning, which will ultimately make for an extraordinary end-
user experience.
Login VSI and Nutanix worked together on this validated reference architecture to demonstrate the performance
benefits provided by Nutanix HCI. Login VSI has become an industry standard for measuring digital workspace
performance and judging by the results of our joint testing, Nutanix delivers outstanding performance with great
efficiency for both XenDesktop and XenApp workspaces. As evidenced by the very low Login VSI baseline of
between 597 & 703ms performance latency, we know that end-users will receive a very responsive desktop
experience, which we also see in the storage latency—with single digit response times in milliseconds. VSImax is
a measure of system saturation, and this reference architecture demonstrates that even when all desktops were
under load, the system still had plenty of headroom, not even reaching the VSImax threshold. Basically, this means
that at no point did the end-user experience degrade to an unacceptable level—a goal all digital workspace
administrators strive for. One of the main contributors to this performance was something unique to Nutanix,
which is AHV, their Acropolis hypervisor. Take a close look… this reference architecture didn’t use XenServer or
VMware ESXi, and the desktop responsiveness, as demonstrated by this white paper, is easily on par with either.
Pretty cool, right!

Audience
This reference architecture document is a 3rd party validation of Nutanix by Login VSI. The document provides
recommendations for architecting, designing, managing and supporting Nutanix infrastructures for Citrix XenApp
and XenDesktop. This document expects its readers to be familiar with Nutanix Acropolis, Prism, AHV and Citrix
XenDesktop.
We have organized this document in such a way as to address the key items for each role that focuses on enabling
a successful design, implementation and transition to operation.
Purpose
This document covers the following subject areas:
o Overview of the Nutanix solution
o Overview of Citrix XenDesktop, XenApp and their use cases
o The benefits of Citrix XenDesktop, XenApp on Nutanix
o Benchmarking Citrix XenDesktop and XenApp performance on Nutanix

Nutanix Enterprise Cloud Platform Overview

Nutanix Acropolis Overview
Nutanix delivers a hyper-converged infrastructure solution purpose-built for virtualization and cloud
environments. This solution brings the performance and economic benefits of web-scale architecture to the
enterprise through the Nutanix Enterprise Cloud Platform, which comprises two different product families—
Nutanix Acropolis and Nutanix Prism.
Attributes of this solution include:
o Storage and compute resources hyper-converged on x86 servers
o System intelligence located in software
o Data, metadata and operations fully distributed across entire cluster of x86 servers
o Self-healing to tolerate and adjust to component failures
o API-based automation and rich analytics
Nutanix Acropolis breaks down into three foundational components:

o Distributed Storage Fabric (DSF)
o App Mobility Fabric (AMF)
o AHV
Nutanix Prism provides one-click infrastructure management for virtual environments running on Acropolis.
Acropolis is hypervisor agnostic, supporting two third-party hypervisors—ESXi and Hyper-V—in addition to the
native Nutanix hypervisor, AHV.
Figure 1. Nutanix Enterprise Cloud Platform

Distributed Storage Fabric

The Distributed Storage Fabric (DSF) delivers enterprise data storage as an on-demand service by employing a
highly distributed software architecture. Nutanix eliminates the need for traditional NAS and SAN solutions, while
delivering a rich set of VM-centric software-defined services. Specifically, the DSF handles the data path of such
features as snapshots, clones, high availability, disaster recovery, deduplication, compression and erasure coding.
The DSF operates via an interconnected network of Controller VMs (CVMs) that form a Nutanix cluster—every
node in the cluster has access to data from shared SSD, HDD and cloud resources. The hypervisors and the DSF
communicate using the industry-standard NFS, iSCSI and SMB3 protocols.
App Mobility Fabric

The App Mobility Fabric (AMF) is the Nutanix virtualization solution that allows apps to move across hypervisors.
When VMs move between hypervisors (for example, between VMware ESXi and AHV), administrators can host
production and dev/test environments concurrently on different hypervisors and shift workloads between them
as needed. A distributed, scale-out service that runs inside the CVM on every node within a Nutanix cluster
implements the AMF solution.
AHV
Nutanix ships with a hardened, enterprise-ready hypervisor based on proven open-source technology. Users
manage AHV via the Prism interface, a robust REST API, and an interactive command-line interface called ACLI
(Acropolis CLI). In combination, these tools eliminate the management complexity typically associated with open
source environments and allow out-of-the-box virtualization on Nutanix—all without the licensing fees
associated with other hypervisors.
Nutanix Acropolis Architecture

Acropolis does not rely on traditional NAS or SAN storage or expensive storage network interconnects. It combines
highly dense storage and server compute (CPU and RAM) into a single platform building block. Each building block
is based on industry-standard Intel processor technology and delivers a unified, scale-out, shared-nothing
architecture with no single points of failure.
The Nutanix solution has no LUNs to manage, no RAID groups to configure, and no complicated storage
multipathing to set up. All storage management is VM-centric, and the DSF optimizes I/O at the VM virtual disk
level. There is one shared pool of storage that includes flash-based SSDs for high performance and HDDs for
affordable capacity. The file system automatically tiers data across different types of storage devices using
intelligent data placement algorithms. These algorithms make sure that the most frequently used data is available
in memory or in flash for the fastest possible performance.

Figure 2. Information Life Cycle Management
The figure below shows an overview of the Nutanix architecture, including the hypervisor of your choice (AHV,
ESXi, or Hyper-V), user VMs, the Nutanix storage CVM and its local disk devices. Each CVM connects directly to
the local storage controller and its associated disks. Using local storage controllers on each host localizes access
to data through the DSF, thereby reducing storage I/O latency. The DSF replicates writes synchronously to at least
one other Nutanix node in the system, distributing data throughout the cluster for resiliency and availability.
Replication factor 2 creates two identical data copies in the cluster, and replication factor 3 creates three identical
data copies. Having a local storage controller on each node ensures that storage performance, as well as storage
capacity, increases linearly with node addition(s).
Figure 3. Overview of the Nutanix Architecture
Local storage for each Nutanix node in the architecture appears to the hypervisor as one large pool of shared
storage. This allows the DSF to support all key virtualization features. Data localization maintains performance
and quality of service (QoS) on each host, minimizing the effect noisy VMs have on their neighbors’ performance.
This functionality allows for large, mixed-workload clusters that are more efficient and more resilient to failure
when compared to traditional architectures with standalone, shared and dual-controller storage arrays.
When VMs move from one hypervisor to another, such as during live migration and high availability, the now local
CVM serves a recently migrated VMs data. When reading old data (stored on the now remote CVM) the local CVM

forwards the I/O request to the remote CVM. All write I/O occurs locally. The DSF detects that the I/O is occurring
from a different node and migrates the data to the local node in the background, which lets a local machine serve
all the read I/O.
The next figure shows how data follows the VM as it moves between hypervisor nodes.
Figure 4. Data Locality and Live Migration
Nutanix Acropolis File Services

Acropolis File Services (AFS) is a scale-out approach that provides Server Message Block (SMB) file services to
Windows clients using SMB 2.1. Acropolis file servers consist of a set of VMs (called FSVMs) on top of the
hypervisor. AFS requires at least three FSVMs running on three nodes to satisfy a quorum for high availability.
Figure 5. Acropolis File Servers Run as VMs to Isolate from DSF
Acropolis File Services (AFS) delivers on-demand performance and automated provisioning to provide highly
scalable file management. It reduces the administration and configuration time needed to deploy and maintain
your environment, providing a public cloud experience within your private cloud.
The Acropolis DSF and Prism make hosted file services highly resilient and easy to use—for example, you can
configure the built-in data efficiency features of DSF, such as compression, deduplication, and EC-X, for each
individual file server. You can also use the Prism UI to administer network and resource management, Active
Directory, fault tolerance and share creation all in one place—vastly improving operational efficiency.
Citrix XenDesktop and XenApp on Nutanix

The Nutanix modular web-scale architecture lets you start small and then expand to meet increasing demand—a
node, a block, or multiple blocks at a time—with no impact on performance. This design removes the hurdle of a
large initial infrastructure purchase, generating faster time-to-value for your XenDesktop and XenApp
implementation.
The figure below presents a typical XenDesktop / XenApp deployment on Nutanix.
Figure 6. XenDesktop on Nutanix: Conceptual Architecture

Solution Design
In the following section, we cover the design decisions and rationale for XenDesktop deployments on Nutanix.
Table 1. General Solution Design Decisions
Item Detail Rationale
Software Version Citrix XenDesktop 7.12
Software Version Acropolis Operating System (AOS) 5.0.2
6 Nutanix Acropolis hosts running AHV

Minimum Size Minimum size requirement
(NX-3000: 2 blocks)
Allow for growth from PoC (hundreds of

Scale Approach Incremental modular scale desktops) to massive scale (thousands of
desktops)
Scale granularly to precisely meet capacity
Scale Unit Node(s), block(s), or pod(s) demands
Scale in n x node increments
Table 2. Nutanix AHV Solution Design Decisions
As many as 32 hosts (starting

Cluster Size Isolated fault domains best practice
with a minimum of 3 hosts)
1x Nutanix DSF datastore per Nutanix handles I/O distribution and
Datastore(s) pod (XenDesktop Server VMs, localization
VM clones, and so on) n-Controller model
Small deployments: Shared

cluster Dedicated infrastructure cluster for
Infrastructure Services
Large deployments: Dedicated larger deployments (best practice)
cluster

Table 3. Nutanix Enterprise Cloud Platform Solution Design Decisions
Cluster Size As many as 32 nodes Isolated fault domains best practice
1x storage pool (SSD, SATA SSD, Standard practice

Storage Pool(s)
SATA HDD) Intelligent tiering handles data locality
1x container for VMs
Container(s) 1x container for data (not used Standard practice
here)
Increase CVM memory to 32 GB
Turn on deduplication and

Features / compression for persistent Needs 32 GB of RAM to enable
Enhancements desktops deduplication
Turn on compression only for

non-persistent desktops
Table 4. Citrix XenDesktop Solution Design Decisions
Minimum: 2 (n+1)
XenDesktop Controllers HA for XenDesktop Controllers
Scale: 1 per additional pod
Users per Controller Up to 5,000 users XenApp/XenDesktop best practice
Ensures availability of controllers

Citrix NetScaler (including
Load Balancing Balances load between controllers and
NetScaler VPX)
pods
Nutanix plug-in for controlling

Desktop Power Management Allows clean reboot and power down
desktop power

Table 5. Citrix StoreFront Solution Design Decisions
StoreFront Servers Minimum: 2 (n+1) HA for StoreFront servers
Ensures availability of StoreFront servers

Citrix NetScaler (including
Load Balancing Balances load between StoreFront
NetScaler VPX)
servers
Table 6. Citrix NetScaler including NetScaler VPX (if used) Solution Design Decisions
NetScaler Servers Minimum: 2 HA for NetScaler (active/passive)
Users per NetScaler Server See product data sheet Varies per model
Ensures availability of NetScaler servers
Load Balancing NetScaler HA Balances load between NetScaler

servers and pods
The next table shows a high-level summary of the pod design for Citrix XenDesktop or XenApp on Nutanix.
Table 7. Control Pod Highlights
Item Quantity
# of XenDesktop/XenApp Controller(s) 2
# of XenDesktop/XenApp StoreFront Server(s) 2
Table 8. Service Pod Highlights
Item Quantity
# of AHV Hosts Up to 32
# of Nutanix Cluster(s) 1
# of Datastore(s) 1

Figure 7. XenDesktop Pod Overview
The section below describes desktop sizing considerations for hosted virtual and streamed desktops.
The following table contains examples of typical scenarios for desktop deployment and use (drawn from Login VSI
definitions).

Table 9. Desktop Scenarios: Typical Definition
Scenario Definition
Task workers and administrative workers perform

repetitive tasks within a small set of applications, usually
at a stationary computer. The applications are usually not
as CPU- and memory-intensive as the applications used
Task Workers
by knowledge workers. Task workers who work specific
shifts might all login to their virtual desktops at the same
time. Task workers include call center analysts, retail
employees and warehouse workers.
Knowledge workers’ daily tasks include accessing the

Internet, using email, and creating complex documents,
Knowledge Workers presentations and spreadsheets. Knowledge workers
include accountants, sales managers and marketing
research analysts.
Power users include application developers and people

Power Users
who use graphics-intensive applications.
The table below proposes initial sizing recommendations for a Windows 7 desktop for these typical scenarios.
These are recommendations only and should be modified after a current state analysis.
Table 10. Desktop Scenario Sizing
Scenario vCPU Memory Disks
Task Workers 1 1.5 GB 35 GB (OS)
Knowledge Workers 2 3 GB 35 GB (OS)
Power Users 2 4 GB 35 GB+ (OS)
Desktop Optimizations
We generated our design with the following high-level desktop guidelines in mind:
o Size desktops appropriately for each specific use case
o Use a mix of applications installed in gold images and application virtualization, depending on the
scenario
o Disable unnecessary OS services and applications
For more detail on desktop optimizations, please refer to the Citrix XenDesktop Windows 10 Optimization Guide.

Citrix and Windows Optimization Best Practices

Windows 10 includes the concepts of versions and branches, which are essential to your Windows 10
implementation. Picking the right version and branch determines your update and validation strategy, which can
directly affect your user experience.
There are multiple versions of Windows 10. For this document, we only address versions applicable to your
XenDesktop implementation: Professional, Education and Enterprise. We do not address Windows 10 Core IoT
(specifically for Internet of Things features running on a Raspberry PI or Arduino) and Home (no domain join).
Microsoft has published a guide to help you choose the right version for your environment.
With the release of Windows 10, Microsoft also introduced the notion of branch releases, which enables Windows
as a Service. This model focuses on adding new capabilities and updates by releasing software in a more pre-
defined and predictable cadence. This delivery model will eventually replace the traditional Windows upgrade
projects, as Windows 10 will get updated to the latest version within the branch without any additional
administrative overhead.
There are four types of branches:
o Windows Insider
o Current Branch (CB)
o Current Branch for Business (CBB)
o Long Term Servicing Branch (LTSB)
The CB branch pushes feature updates when Microsoft releases them. With Windows 10 version 1607, which
Microsoft recommends, you can use the servicing tools to delay CB feature updates up to 180 days before
implementation. The CB branch is a good fit for pilot deployments and early adopters that need the latest features
directly.
Where the CB branch suits early adopters by providing a test cycle before the major branch roll out, the CBB
branch is typically deployed for the actual major roll out. The CBB branch gets the same version numbers as the
CB branch, but at a later point in time to make sure that organizations are ready for the wider distribution of the
release.
Systems that require a longer servicing branch due to their specialized use cases benefit from the LTSB branch.
The LTSB targets systems with devices that have limited changes and must be stable and always available, such as
medical equipment or ATMs.
Optimizations
Login VSI has a detailed document on how to optimize Windows 10 for XenDesktop, which has been consolidated
into a template for the VMware OS Optimization Tool. Applying these best practices significantly increases user
and desktop per host ratio density. Login VSI has also shown these optimizations improve VSIMax 44 percent over
a ‘vanilla’ Windows 10 installation. For more detailed information about the optimizations, see this recorded
session.
The Login VSI template for the VMware OS optimization tool, which provides the best density while maintaining
the optimal user experience, handles most of these best practice elements. For this reference architecture, we

followed the Citrix Windows 10 Optimization Guide, BPAnalyzer and the Login VSI Windows 10 template. The main
items of all three sources are:
o Features
o Scheduled tasks
o Services
o Modern apps
o Microsoft OneDrive
o Hardware acceleration
Features
The new features in Windows 10 that have generated the most discussion are those that relate to privacy and
telemetry. When Citrix XenDesktop delivers Windows 10, these features negatively affect performance, while
providing only minimal gains for user experience. Considering this fact, disabling these new features enhances
security and privacy, while also improving user density.
Best Practice: Disabling new features such as boot logging and call home, as well as other
telemetry-related features improves both security and user density.
Scheduled Tasks
Any default Windows client operating system schedules tasks by default. While these scheduled tasks are useful
and have little effect on a physical deployment, they can have a significant effect on your XenDesktop image. The
newly introduced Indexer Automatic Maintenance task, for example, scans your Windows operating system
anytime the machine is idle to check for errors, defragment the hard drive and run optimization services. Running
this task on hundreds or thousands of desktops in a centralized environment negatively affects your system’s
active users. Other scheduled tasks magnify this effect, which is why we recommend disabling scheduled tasks
whenever possible.
Best Practice: When implementing XenDesktop, disable scheduled tasks whenever possible,
including defragmentation, automatic maintenance and Bluetooth scanning.
Services
XenDesktop lets you consolidate your desktop centralization into one or more datacenters, which means that
some of the default services that come with Windows 10 add little benefit. Other default services add limited
value when desktops are virtualized (rather than deployed on PCs or laptops). Services like BranchCache, Internet
Connection Sharing and Geolocation don’t add value when XenDesktop delivers the desktop services from a
central location. Accordingly, disabling these services has the advantage of hardening your operating system and
improving user experience by lowering the operating system overhead.
Best Practice: Disabling unused services hardens your operating system and lowers the
operating system overhead, which improves user experience.

Consumer-focused Apps
The Windows 10 default deployment includes a list of preinstalled applications, many of which are consumer
focused, such as Twitter, Asphalt 8, Airborn and FarmVille 2. Removing these consumer-market applications frees
up resources and helps limit end-user distraction.
Best Practice: Removing consumer-focused apps frees up machine resources and limits end-
user distraction.
Microsoft OneDrive
Microsoft OneDrive is a file synchronization solution that allows you to connect to your Office 365 account (for
example) and store your files both locally and in the Cloud. Microsoft, however, did not build OneDrive for
virtualized desktops, as it synchronizes the complete OneDrive content locally. If your XenDesktop deployment
contains non-persistent desktops, OneDrive syncs the complete folder per user after every new logon to the
system. In accordance, we recommend removing Microsoft OneDrive to avoid wasting disk capacity and to
prevent network congestion.
Best Practice: Uninstalling Microsoft OneDrive eliminates network congestion due to user
folder synchronization actions with Office 365.
Hardware Acceleration
Since the release of Microsoft Office 2013, Office has used a hardware-accelerated method for presenting Office
programs, such as Word and Excel. The Microsoft operating system offers this feature, but the operating system
itself relies on the graphics processing unit (GPU) and the display drivers. If there is no GPU available, the CPU
performs the rendering, which uses additional system resources.
Best Practice: If you have no GPU available, make sure to enable software rendering for
Internet Explorer and to disable hardware acceleration for Office 2010, Office 2013 and Office
2016 to improve the user experience.

AHV Pod Design

The tables below contain highlights from a high-level snapshot of the Citrix XenDesktop on Nutanix Hosted Virtual
Desktop Pod.
Table 11. AHV Control Pod Detail
Item Quantity
# of XenDesktop Controller(s) 2
# of XenDesktop StoreFront
2
Server(s)
Table 12. AHV Services Pod Detail
Item Quantity
# of Datastore(s) 1
# of Desktops Up to 4,800
MCS Pod Design

The following table contains highlights from a high-level snapshot of the Citrix XenApp on Nutanix Hosted Virtual
Desktop Pod.
Table 13. MCS Control Pod Detail
Item Quantity
# of XenApp Controller(s) 2
# of XenApp StoreFront Server(s) 2
Table 14. MCS Services Pod Detail
Item Quantity
# of Datastore(s) 1
# of Desktops Up to 8,000

Nutanix: Compute and Storage

The Nutanix Enterprise Cloud Platform provides an ideal combination of both high-performance compute and
localized storage to meet any demand. True to this capability, this reference architecture contains no optimized
reconfiguration or customization of the Nutanix product for this use case.
The following figure shows a high-level example of the relationship between the Nutanix storage pool and its
containers.
Figure 8. Nutanix Logical Storage Configuration
The table below includes the Nutanix storage pool and container configuration.
Table 15. Nutanix Storage Configuration
Name Role Details
SP01 Main storage pool for all data SSD and HDD
CTRVSI Container for all VMs AHV Datastore
CTR-RF2-DATA-01 Container for all data (not used here) AHV Datastore

Solution Application: XenDesktop

This section applies the recommended pod-based reference architecture to real-world scenarios for XenDesktop
and outlines the sizing metrics and components.
Scenario: 900 Desktops

Table 16. Infrastructure Breakdown: 900 Desktops Using NX-3060-G5
Item Value
# of AHV hosts 6
# of datastore(s) 1
Table 17. Component Breakdown: 900 Desktops Using NX-3060-G5
Item Value
# of Nutanix desktop
1 (partial)
pods
# of Nutanix blocks 2
# of RU (Nutanix) 4
# of 10 GbE ports 16
# of 100/1000 ports (IPMI) 8
# of L2 leaf switches 2
# of L3 spine switches 1

Scenario: 1,200 Desktops

Table 18. Infrastructure Breakdown: 1200 Desktops Using NX-3060-G5
Item Value
# of AHV hosts 8
# of datastore(s) 1
Table 19. Component Breakdown: 1200 Desktops Using NX-3060-G5
Item Value
1 (partial)
pods
# of Nutanix blocks 2
# of RU (Nutanix) 4
# of 10 GbE ports 16
# of 100/1000 ports (IPMI) 8

Solution Application: XenApp

This section applies the recommended pod-based reference architecture to real-world scenarios for XenApp and
outlines the sizing metrics and components.
Scenario: 1,000 Desktops

Table 20. Infrastructure Breakdown: 1,000 Desktops Using NX-3060-G5
Item Value
# of AHV hosts 4
# of datastore(s) 1
Table 21. Component Breakdown: 1,000 Desktops Using NX-3060-G5
Item Value
1 (partial)
pods
# of Nutanix chassis
1
(blocks)
# of RU (Nutanix) 2
# of 10 GbE ports 8
# of 100/1000 ports (IPMI) 4
Validating and Benchmarking

We conducted the solution and testing in this document with Citrix XenDesktop 7.12 deployed on AHV and the
Nutanix Enterprise Cloud Platform. Login VSI testing was used to validate performance of the Nutanix Enterprise
Cloud platform.

Environment Overview
We used two nodes of an existing Nutanix NX-8050 to host all infrastructure and XenDesktop services and a
Nutanix NX-3460 to host the Login VSI test harness. The six nodes in the eight node Nutanix NX-3060-G5
functioned as the target environment and provided all desktop hosting. The Nutanix block connected to an Arista
7050S top-of-rack switch using 10 GbE.
Figure 9. Test Environment Overview

Test Environment Configuration

Assumptions:
o Knowledge Worker use case
o Per-desktop IOPS (Knowledge Worker): 10 sustained and 70 peak (startup)
o Using MCS persistent Machine Catalogs
Hardware:
o Storage and compute: 2x Nutanix NX-3460-G5
o Network: Arista 7050Q (L3 spine) and 7050S (L2 leaf) series switches
Desktop configuration:
o OS: Windows 10 x64
o 2 vCPU and 3 GB RAM
o 1x 32 GB OS Disk
Applications:
o Microsoft Office 2013
o Adobe Acrobat Reader XI
o Internet Explorer
o Flash video
Login VSI:
o Login VSI 4.1 Professional
XenDesktop Configuration
The table below shows the XenDesktop configuration used in the test environment.
Table 22. XenDesktop Configuration
VM Qty vCPU Memory Disks
XenDesktop Controller(s) 1 4 8 1 x 40 GB (OS)
StoreFront Server(s) 1 4 4 1 x 40 GB (OS)
XenApp Configuration
The table below shows the configuration used in the test environment.
Table 23. XenApp Configuration
VM Quantity vCPU Memory Disks
Delivery Controller(s) 1 4 8 1x 40 GB (OS)
StoreFront Server(s) 1 4 4 1x 40 GB (OS)

Login VSI Benchmark

Login VSI provides performance insights for virtualized desktop and server environments. Enterprise IT
departments use Login VSI products in all phases of their virtual desktop deployment—from planning to
deployment to change management—to facilitate more predictable performance, higher availability and a more
consistent end-user experience. The world’s leading virtualization vendors use the flagship product, Login VSI, to
benchmark performance. With minimal configuration, Login VSI products works in VMware Horizon View, Citrix
XenDesktop and XenApp, Microsoft Remote Desktop Services (Terminal Services) and any other Windows-based
virtual desktop solution.
For more information about Login VSI, please visit: www.loginvsi.com
The following table includes all four workloads available in Login VSI 4.1.
Table 24. Login VSI 4.1 Workloads
Task Worker Office Worker Knowledge Worker Power User
Light Medium Medium Heavy
1 vCPU 1 vCPU 2 vCPUs 2–4 vCPUs
2–3 Apps 4–6 Apps 4–7 Apps 5–9 Apps
No video 240p video 360p video 720p video
Login VSI Workloads

For more information on Login VSI workloads, please read this Login VSI blog on the subject.
Login VSI Testing Best Practices

Desktop & Server Optimizations
Tuning your desktop and servers with respect to the user experience; it is important to understand the value of
tuning your templates as it will directly affect the user experience and density. It is a delicate balance to tune
the image for density while retaining the user experience.
An Example – If you were to disable windows search services your users would lose this ability within their
desktop sessions. This would reduce the machine foot print as far as processor, disk IO, etc. at the expense of
user experience.
We covered the tuning utilized in this reference architecture within the “Citrix and Windows Optimization Best
Practices” section.

Power Configuration for Performance

Modern day hardware is configured to protect the environment, and scale the power consumption based upon
the load on the current system. While this configuration is ideal for cutting energy consumption costs it will lead
to inconsistent test results. Due to this we recommend that testing is performed while hyper visor hardware is
configured for “High Performance”. The method for changing this differs between hardware providers.
Note – Ensure that “High Performance” is configured on the Hypervisor, and within the Windows OS.
Use Flash Storage for User Profiles

The speed at which a user gets their profile information directly effects their user experience. It is important
when you are dealing with user created customization of their user experience that the response is fast. We
have seen many instances of user profile content being stored on a slow NAS device. Utilizing an appropriate
mix of storage devices should assist with addressing this bottleneck.
How to Interpret the Results

Login VSI
Login VSI is a test benchmark used to simulate real-world user workload on a desktop. These values represent the
time it takes for an application or task to complete (launching Outlook, for example) and is not in addition to
traditional desktop response times. These figures do not refer to the round-trip time (RTT) for network I/O—they
refer to the total time to perform an action on the desktop.
During the test, all VMs turn on and the workload starts on a new desktop every 30 seconds, until all sessions and
workloads are active.
We quantified the evaluation using the following metrics:
o Minimum Response (the minimum application response time)
o Average Response (the average application response time)
o Maximum Response (the maximum application response time)
o VSI Baseline (average application response time of the first 15 sessions)
o VSI Index Average (the average response time, dropping the highest and lowest two percent)
o VSImax (reached at the first session to record a VSI Timing metric 1000ms beyond the VSI Baseline)

Results: Citrix XenDesktop on Nutanix
900 Knowledge Worker Desktops (Six Nodes)

During testing with 900 desktops, the performance did not reach the VSImax threshold. The test produced a VSI
Index Average of 1,302 ms, a VSImax of 1,703 ms and a VSIbase of 703 ms.
Figure 60. 900 Knowledge Worker Desktops

Figure 11. Peak CPU Utilization

1200 Knowledge Worker Desktops (Eight Nodes)

During testing with 1200 desktops, the performance did not reach the VSImax threshold. The test produced a VSI
Figure 12. - 1200 Knowledge Worker Desktops
CPU utilization for the AHV hosts during this test peaked at 88.73 percent.

Nutanix Metrics
IOPS peaked at approximately 17,927 during the high-volume startup period to refresh the desktops; the peak
IOPS value during the tests was 6,510.
Figure 13. Peak IOPS
Command latency peaked at approximately 5.55 ms during the tests, as shown in the figure below.
Figure 14. Command Latency Peak

Results: Citrix XenApp on Nutanix

1,000 Knowledge Worker Desktops (Four Nodes)
During testing with 1,000 desktops, the performance did not reach the VSImax threshold. The test produced a VSI
Figure 75. 1,000 Knowledge Worker Desktops

Nutanix Metrics
IOPS peaked at approximately 3,926 during the high-volume startup period to refresh the desktops; the peak IOPS
value during the tests was 2,853.
Figure 16.8 Peak IOPS
Command latency peaked at approximately 5.66 ms during the tests, as shown in the figure below.
Figure 17. Command Latency Peak
CPU usage peaked at approximately 98.67 during the tests, as shown in the figure below.
Figure 18. CPU Usage Peak

Conclusion
Our extensive testing of Citrix XenDesktop on Nutanix revealed a light I/O footprint on the Nutanix Enterprise
Cloud Platform. Aggregate IOPS peaked at approximately 18,000 during the high-volume startup periods, and
sustained IOPS ranged from 5,000–6,000. I/O latencies averaged less than 2 ms for read and less than 5 ms for
write, during peak load.
Our extensive testing of Citrix XenApp on Nutanix also revealed a light I/O footprint on the Nutanix Enterprise
Cloud Platform. Aggregate IOPS peaked at approximately 4,000 during the high-volume startup periods, and
sustained IOPS ranged from 2,000-3,000. I/O latencies averaged less than 2 ms for read and less than 5 ms for
write, during peak load.
In conclusion, Nutanix offers the performance and low latency necessary for an uncompromised end-user
experience—even during boot storms, patch and update operations, and application and OS upgrades.
The Citrix XenDesktop-on-Nutanix solution provides a single, high-density platform for desktop and application
delivery.
Running XenDesktop on Nutanix means that you can:
o Get started quickly, without onerous upfront expenditures, and then scale as you grow
o Deliver excellent user density, supporting 150 Knowledge Workers per AHV host
o Simplify management and administration by eliminating LUNs
o Significantly lower acquisition and ongoing costs by running VDI alongside other workloads
While running XenApp on Nutanix means that you can:

o Get started quickly, without onerous upfront expenditures, and then scale as you grow
o Deliver excellent user density, supporting 250 Knowledge Workers per AHV host
o Simplify management and administration by eliminating LUNs
o Significantly lower acquisition and ongoing costs by running SBC alongside other workloads

Appendix
Configuration
Hardware:
o Storage and compute: Nutanix NX-3460-G5
o Per-node specs (four nodes per 2RU block):
o CPU: 2x Intel Xeon E5-2680v4 @ 2,4Ghz
o RAM: 512 GB
o Network:
o Arista 7050Q L3 spine
o Arista 7050S L2 leaf
Software:
o Nutanix Acropolis: AOS 5.0.2
o XenDesktop: 7.12
o Desktop:
o Windows 10 x64
o Office 2013
o Infrastructure: AHV
VM (XenDesktop):
o Desktop
o CPU: 2 vCPU
o Memory: 3 GB RAM
o Storage: 1x 32 GB OS Disk on CTRVSI DSF datastore
VM (XenApp):
o CPU: 8 vCPU
o Memory: 32 GB RAM
o Storage: 1x 50 GB OS Disk on CTRVSI DSF datastore

Performance
Nutanix has identified a potential performance bottleneck due to a bug in QEMU and USB-tablet CPU
consumption. The USB host adapter that polls the USB devices may use excessive CPU resources when emulating
the host adapter. While already solved for Linux-based operating systems, this problem still affects Windows,
which impacts your XenDesktop implementation.
Nutanix plans to create a fix for the newer releases of their Acropolis operating system (AOS). For VMs created on
AOS 4.7.3 or earlier, an in-guest fix can bring down CPU utilization and increase performance.
Once you confirm that USB type 1 devices are in use, delete these two registry subtrees:
HLM\SYSTEM\CurrentControlSet\Control\usbflags
HLM\SYSTEM\CurrentControlSet\Enum\USB
After rebooting the machine, Windows checks the OS descriptors again; at this point the CPU overhead fix is now
in place.
About Login VSI

Login VSI provides predictive and advanced insights for virtualized desktop and server based computing
environments. Our software ensures Enterprises can provide their employees with the optimum end-user
experience. By utilizing virtual users that mimic real-world users performing real-world tasks— predictive analysis
will increase productivity while minimizing downtime and other costly business disruptions. With minimal
configuration, Login VSI products work in VMware Horizon®, Citrix XenDesktop and XenApp, Microsoft Remote
Desktop Services (Terminal Services) and any other Windows-based virtual desktop solution. Visit
http://www.loginvsi.com for more information.

Reference Architecture - Citrix On Nutanix AHV

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Reference Architecture - Citrix On Nutanix AHV

Uploaded by

Copyright:

Available Formats

Citrix on Nutanix AHV

Disclosure and Warranty

Validated by Login VSI 1

1.0 Document Creation Brian Martynowicz May 16, 2017

About the Author

Validated by Login VSI 2

Validated by Login VSI 3

Citrix XenApp on Nutanix AHV Summary

Citrix XenDesktop on Nutanix AHV Summary

Validated by Login VSI 4

Validated by Login VSI 5

Validated by Login VSI 6

Nutanix Enterprise Cloud Platform Overview

Nutanix Acropolis breaks down into three foundational components:

Figure 1. Nutanix Enterprise Cloud Platform

Validated by Login VSI 7

Distributed Storage Fabric

App Mobility Fabric

Nutanix Acropolis Architecture

Validated by Login VSI 8

Figure 2. Information Life Cycle Management

Figure 3. Overview of the Nutanix Architecture

Validated by Login VSI 9

Figure 4. Data Locality and Live Migration

Nutanix Acropolis File Services

Figure 5. Acropolis File Servers Run as VMs to Isolate from DSF

Citrix XenDesktop and XenApp on Nutanix

Figure 6. XenDesktop on Nutanix: Conceptual Architecture

Validated by Login VSI 11

Item Detail Rationale

Software Version Citrix XenDesktop 7.12

Software Version Acropolis Operating System (AOS) 5.0.2

6 Nutanix Acropolis hosts running AHV

Allow for growth from PoC (hundreds of

Scale granularly to precisely meet capacity

Scale Unit Node(s), block(s), or pod(s) demands

Scale in n x node increments

Table 2. Nutanix AHV Solution Design Decisions

Item Detail Rationale

As many as 32 hosts (starting

1x Nutanix DSF datastore per Nutanix handles I/O distribution and

Datastore(s) pod (XenDesktop Server VMs, localization

VM clones, and so on) n-Controller model

Small deployments: Shared

Validated by Login VSI 12

Table 3. Nutanix Enterprise Cloud Platform Solution Design Decisions

Item Detail Rationale

Cluster Size As many as 32 nodes Isolated fault domains best practice

1x storage pool (SSD, SATA SSD, Standard practice

1x container for VMs

Container(s) 1x container for data (not used Standard practice

Increase CVM memory to 32 GB

Turn on deduplication and

Enhancements desktops deduplication

Turn on compression only for

Table 4. Citrix XenDesktop Solution Design Decisions

Item Detail Rationale

Users per Controller Up to 5,000 users XenApp/XenDesktop best practice

Ensures availability of controllers

Nutanix plug-in for controlling

Validated by Login VSI 13

Table 5. Citrix StoreFront Solution Design Decisions

Item Detail Rationale