Professional Documents
Culture Documents
Reference Architecture - Citrix On Nutanix AHV
Reference Architecture - Citrix On Nutanix AHV
By Brian Martynowicz
Citrix on Nutanix AHV
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.
Revision History
Revision Change Description Updated By Date
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.
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 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.
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.
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).
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.
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.
Validated by Login VSI 10
Citrix on Nutanix AHV
Solution Design
In the following section, we cover the design decisions and rationale for XenDesktop deployments on Nutanix.
Table 1. General Solution Design Decisions
cluster
here)
Minimum: 2 (n+1)
XenDesktop Controllers HA for XenDesktop Controllers
Scale: 1 per additional pod
Table 6. Citrix NetScaler including NetScaler VPX (if used) Solution Design Decisions
Users per NetScaler Server See product data sheet Varies per model
The next table shows a high-level summary of the pod design for Citrix XenDesktop or XenApp on Nutanix.
Item Quantity
# of XenDesktop/XenApp Controller(s) 2
Item Quantity
# of AHV Hosts Up to 32
# of Nutanix Cluster(s) 1
# of Datastore(s) 1
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).
Scenario Definition
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
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.
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.
Item Quantity
# of XenDesktop Controller(s) 2
# of XenDesktop StoreFront
2
Server(s)
Item Quantity
# of AHV Hosts Up to 32
# of Nutanix Cluster(s) 1
# of Datastore(s) 1
# of Desktops Up to 4,800
Item Quantity
# of XenApp Controller(s) 2
Item Quantity
# of AHV Hosts Up to 32
# of Nutanix Cluster(s) 1
# of Datastore(s) 1
# of Desktops Up to 8,000
The table below includes the Nutanix storage pool and container configuration.
Table 15. Nutanix Storage Configuration
SP01 Main storage pool for all data SSD and HDD
CTR-RF2-DATA-01 Container for all data (not used here) AHV Datastore
Item Value
# of AHV hosts 6
# of datastore(s) 1
Item Value
# of Nutanix desktop
1 (partial)
pods
# of Nutanix blocks 2
# of RU (Nutanix) 4
# of 10 GbE ports 16
# of L2 leaf switches 2
# of L3 spine switches 1
Item Value
# of AHV hosts 8
# of datastore(s) 1
Item Value
# of Nutanix desktop
1 (partial)
pods
# of Nutanix blocks 2
# of RU (Nutanix) 4
# of 10 GbE ports 16
# of L2 leaf switches 2
# of L3 spine switches 1
Item Value
# of AHV hosts 4
# of datastore(s) 1
Item Value
# of Nutanix desktop
1 (partial)
pods
# of Nutanix chassis
1
(blocks)
# of RU (Nutanix) 2
# of 10 GbE ports 8
# of L2 leaf switches 2
# of L3 spine switches 1
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.
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
XenApp Configuration
The table below shows the configuration used in the test environment.
Table 23. XenApp Configuration
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.
Note – Ensure that “High Performance” is configured on the Hypervisor, and within the Windows OS.
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.
Command latency peaked at approximately 5.55 ms during the tests, as shown in the figure below.
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.
Command latency peaked at approximately 5.66 ms during the tests, as shown in the figure below.
CPU usage peaked at approximately 98.67 during the tests, as shown in the figure below.
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
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.