White Paper - Citrix XenDesktop on vSphere With Nutanix

Citrix XenDesktop on vSphere Reference Architecture

v1.0 – March 2013

2

Citrix XenDesktop on vSphere | 2

| 2

Copyright 2013 Nutanix, Inc.

All rights reserved. This product is protected by U.S. and international copyright and intellectual property laws.

Nutanix is a trademark of Nutanix, Inc. in the United States and/or other jurisdictions. All other marks and names mentioned herein may be trademarks of their respective companies.

3


| 3

Table of Contents

1. Executive Summary ..................................................................................... 5

2. Introduction .................................................................................................. 6

3. Solution Overview ........................................................................................ 7

3.1. What is the Nutanix Architecture? .......................................................................... 7

3.2. What is Citrix XenDesktop? ................................................................................... 8

3.3. Citrix XenDesktop the Nutanix Way ..................................................................... 11

4. Solution Design .......................................................................................... 13

4.1. XenDesktop Machine Creation Services (MCS) .................................................. 17

4.2. XenDesktop Provisioning Services (PVS) ........................................................... 20

4.3. Nutanix – Compute/Storage ................................................................................. 24

4.4. Network ................................................................................................................ 25

5. Solution Application ................................................................................... 26

5.1. Scenario: 300 Desktops ....................................................................................... 26

5.2. Scenario: 600 Desktops ....................................................................................... 27

5.3. Scenario: 1,200 Desktops .................................................................................... 28




5.7. Scenario: 19,200 Desktops .................................................................................. 32

6. Validation & Benchmarking ........................................................................ 33

6.1. Environment Overview ......................................................................................... 33

6.2. Login VSI Benchmark .......................................................................................... 37

6.3. How to Interpret the Results ................................................................................ 39

7. Results ....................................................................................................... 42

7.1. MCS – 400 Light Desktops .................................................................................. 42

4


| 4

7.2. MCS – 300 Medium Desktops ............................................................................. 44

7.3. PVS – 400 Light Desktops ................................................................................... 46

7.4. PVS – 300 Medium Desktops .............................................................................. 49

7.5. PVS – 360 Medium Desktops .............................................................................. 52

8. Further Research ....................................................................................... 55

9. Conclusion ................................................................................................. 56

10. Appendix: Configuration ............................................................................. 57

11. References ................................................................................................ 59

11.1. Table of Figures ................................................................................................... 59

11.2. Table of Tables .................................................................................................... 60

12. About the Author ........................................................................................ 61

5


| 5

1. Executive Summary

This document makes recommendations for the design, optimization and scaling of Citrix XenDesktop deployments on Nutanix. It shows the scalability of the Nutanix virtual computing platform and provides detailed performance and configuration information on the scale-out capabilities of the cluster when leveraged for XenDesktop deployments.

Extensive testing has been performed using Login VSI to simulate real-world workloads and conditions of a XenDesktop environment using MCS and PVS on Nutanix. The sizing data and recommendations made in this document are based upon multiple testing iterations and thorough technical validation.

The solution and testing data was completed with Citrix XenDesktop deployed on VMware vSphere, both running on the Nutanix virtual computing platform. Testing validation was done using Login VSI.

Desktop user density will be primarily driven by the available host CPU resources and not due to any IO or resource bottleneck for both MCS and PVS deployments on Nutanix. Login VSI Light test results showed densities of over 400 task worker desktops per 2U Nutanix appliance. However, most VDI deployments fit within the knowledge worker category which was validated at over 300 desktops per 2U appliance using the Login VSI Medium test. Sizing for the pods was determined after careful consideration of performance as well as accounting for additional resources for N+1 failover capabilities.

6


| 6

2. Introduction

Audience

This reference architecture document is part of the Nutanix Solutions Library and is intended for architecting, designing, managing, and/or supporting Nutanix infrastructures. Consumers of this document should be familiar with VMware vSphere, Citrix XenDesktop, and Nutanix.

We have broken down this document to address key items for each role focusing on the enablement of a successful design, implementation, and transition to operation.

Purpose

This document will cover the following subject areas:

o Overview of the Nutanix solution

o Overview of Citrix XenDesktop and its use cases

o The benefits of Citrix XenDesktop on Nutanix

o Architecting a complete Citrix XenDesktop solution on the Nutanix platform

o Design and configuration considerations when architecting a Citrix XenDesktop solution on Nutanix

o Benchmarking Citrix XenDesktop performance on Nutanix

7


| 7

3. Solution Overview

3.1. What is the Nutanix Architecture?

Nutanix virtual computing platform is a scale-out cluster of high-performance nodes, or servers, each running a standard hypervisor and containing processors, memory and local storage (consisting of PCIe SSD Flash and high capacity SATA disk drives). Each node runs virtual machines just like a standard virtual machine host. In addition, local storage from all nodes is virtualized into a unified pool by the Nutanix Distributed File System (NDFS) (Figure 1). In effect, NDFS acts like an advanced NAS that uses local SSDs and disks from all nodes to store virtual machine data. Virtual machines running on the cluster write data to NDFS as if they were writing to shared storage. NDFS is VM aware and provides advanced data management features. It brings data closer to virtual machines by storing the data locally on the system, resulting in higher performance at a lower cost. Nutanix Complete Cluster can horizontally scale from as few as three nodes to a large number of nodes, enabling organizations to scale their infrastructure as their needs grow.

Figure 1 Nutanix Architecture

Inspired by the Google File System, NDFS delivers a unified pool of storage from all nodes across the cluster, leveraging techniques including striping, replication, auto-tiering, error detection, failover’ and automatic recovery. This pool can then be presented as shared storage resources to VMs for seamless support of features like vMotion, HA, and DRS, along with industry-leading data management features. Additional nodes can be added in a plug-and-play manner in this high-performance scale-out architecture to build a cluster that will easily grow as your needs do.

8


| 8

3.2. What is Citrix XenDesktop?

Citrix XenDesktop is a desktop virtualization solution that transforms desktops and applications into a secure on-demand service available to any user, anywhere, on any device. With XenDesktop, you can deliver individual Windows, web and SaaS applications, or full virtual desktops, to PCs, Macs, tablets, smartphones, laptops and thin clients with a high-definition user experience.

Citrix XenDesktop provides a complete virtual desktop delivery system by integrating several distributed components with advanced configuration tools that simplify the creation and real-

time management of the virtual desktop infrastructure.

The core components of XenDesktop are:

o Desktop Delivery Controller. Installed on servers in the data center, the controller authenticates users, manages the assembly of users' virtual desktop environments, and brokers connections between users and their virtual desktops. It controls the state of the desktops, starting and stopping them based on demand and administrative configuration. Desktop Delivery Controller also includes Profile management, in some editions, to manage user personalization settings in virtualized or physical Windows environments.

o Virtual Desktop Provisioning powered by Citrix Provisioning Services. Provisioning Services creates and provisions virtual desktops from a single desktop image on demand, optimizing storage utilization and providing a pristine virtual desktop to each user every time they log on. Desktop provisioning also simplifies desktop images, provides the best flexibility, and offers fewer points of desktop management for both applications and desktops.

o Virtual Desktop Agent. Installed on virtual desktops, the agent enables direct ICA (Independent Computing Architecture) connections between the virtual desktop and user devices.

o Citrix Receiver. Installed on user devices, the Citrix Desktop Receiver enables direct ICA connections from user devices to virtual desktops.

o Citrix FlexCast. Citrix XenDesktop with FlexCast delivery technology lets you deliver virtual desktops and applications tailored to meet the diverse performance, security and flexibility requirements of every worker in your organization through a single solution. Centralized, single-instance management helps you deploy, manage and secure user desktops more easily and efficiently.

9


| 9

Deployment Scenario – Machine Creation Services (MCS)

Machine Creation Services: Provides images only to desktops virtualized on a hypervisor. The images are contained within the hypervisor pool and then thin provisioned as needed. The thin provisioned virtual desktops utilize identity management functionality to overcome the new security identity (SID) requirements typical with cloning. Machine Creation Services is managed by the XenDesktop Controllers and utilizes the capabilities of the underlying hypervisor.

MCS does not require additional servers; it utilizes integrated functionality built into Citrix XenServer, Microsoft Hyper-V and VMware vSphere. As MCS utilizes hypervisor functionality, it is only a viable option for desktops virtualized on a hypervisor. A master desktop image is created and maintained within the hypervisor pool. The XenDesktop Controller instructs the hypervisor to create a snapshot of the base image and thin provision new virtual machines through the built-in hypervisor functions. However, thin provisioning images often results in cloning issues as each provisioned desktop has the same identity as the master. MCS utilizes special functionality within the XenDesktop Controller and XenDesktop Agent (installed within the virtual desktop image) to build unique identities for each virtual machine, which is stored within the virtual desktops identity disk. This functionality allows each virtual desktop to be unique even though it is using the same base image

Figure 2 Machine Creation Services

10


| 10

Deployment Scenario – Provisioning Services (PVS)

Provisioning Services streaming technology allows computers to be provisioned and re-provisioned in real-time from a single shared-disk image. In doing so, administrators can completely eliminate the need to manage and patch individual systems. Instead, all image management is done on the master image. The local hard-disk drive of each system may be used for runtime data caching or, in some scenarios, removed from the system entirely, which reduces power usage, system failure rates, and security risks. Provisioning Services can stream these images to both virtual and physical devices.

The Provisioning Services solution’s infrastructure is based on software-streaming technology. After installing and configuring Provisioning Services components, a vDisk is created from a device’s hard drive by taking a snapshot of the OS and application image, and then storing that image as a vDisk file on the network. A device that is used during this process is referred to as a Master target device. The devices that use those vDisks are called target devices.

vDisks can exist on a Provisioning Server, file share, or in larger deployments, on a storage system that the Provisioning Server can communicate with (iSCSI, SAN, NAS, and CIFS). vDisks can be assigned to a single target device as Private Image Mode, or to multiple target devices as Standard Image Mode.

When a target device is turned on, it is set to boot from the network and to communicate with a Provisioning Server. Unlike thin-client technology, processing takes place on the target device (refer to Step 1 in the illustration that follows).

Figure 3 Provisioning Services

11


| 11

3.3. Citrix XenDesktop the Nutanix Way

The Nutanix platform operates and scales Citrix XenDesktop MCS and PVS. Figure 4 shows the XenDesktop on Nutanix solution:

Desktop BrokersDesktop Brokers

Hosted Desktops PoolsHosted Desktops Pools

Dedicated/Pooled/Shared w/MCS

Infrastructure Services


Active Directory

vCenter

SiteSite

User Data

XenDesktop Controller Server Pool

Site Load Balancer(s)Site Load Balancer(s)

SCA

LE

vCOps

Master vDisks Provisioning Services

Streamed Desktops PoolsStreamed Desktops Pools

Master Images

Streamed w/ PVS

SCA

LE

StorefrontStorefront

Figure 4 XenDesktop on Nutanix Conceptual Arch

The Nutanix approach of modular scale-out enables customers to select any initial deployment size and grow in more granular data and desktop increments. This removes the hurdle of a large up-front infrastructure purchase that a customer will need many months or years to grow into, ensuring a faster time-to-value for the a XenDesktop implementation.

The Nutanix solution is fully integrated with the VMware APIs for Array Integration (VAAI) and provides high performance PCIe SSD flash to enable you to provide the best possible experience to the end user with the flexibility of a single modular platform.

12


| 12

Why run Citrix XenDesktop on Nutanix?

Nutanix enables you to run multiple workloads all on the same scalable converged infrastructure

o Modular incremental scale: With the Nutanix solution you can start small and scale. A single Nutanix block provides up to 20TB storage and 400 desktops in a compact 2U footprint. Given the modularity of the solution, you can granularly scale per-node (up to ~5TB/100 desktops), per-block (up to ~20TB/400 desktops), or with multiple blocks giving you the ability to accurately match supply with demand and minimize the upfront CapEx.

o Integrated: The Nutanix platform provides full support for VAAI allowing you to leverage all the latest advancements from VMware and taking your VDI solution to the next level.

o High performance: Up to 25,000 random IOPS and up to 2,000 MB/s of sequential throughput in a compact 2U 4-node cluster.

o Change management: Maintain environmental control and separation between development, test, staging, and production environments. Snapshots and fast clones can help in sharing production data with non-production jobs, without requiring full copies and unnecessary data duplication.

o Business continuity and data protection: User data and desktops are mission critical and need enterprise-grade data management features including backup and DR. With Nutanix these are provided out-of-the-box and can be managed the same as they would be for virtual environments.

o Data efficiency: The Nutanix solution is truly VM-centric for all compression policies. Unlike traditional solutions that perform compression mainly at the LUN level, the Nutanix solution provides all of these capabilities at the VM and file level, greatly increasing efficiency and simplicity. These capabilities ensure the highest possible compression/decompression performance on a sub-block level. By allowing for both inline and post-process compression capabilities, the Nutanix solution breaks the bounds set by traditional compression solutions.

o Enterprise-grade cluster management: A simplified and intuitive Apple-like approach to managing large clusters, including a converged GUI that serves as a single pane of glass for servers and storage, alert notifications, and bonjour mechanism to auto-detect new nodes in the cluster. Spend more time enhancing your environment, not maintaining it.

o High-density architecture: Nutanix uses an advanced server architecture in which 8 Intel CPUs (up to 64 cores) and up to 1TB of memory are integrated into a single 2U appliance. Coupled with data archiving and compression, Nutanix can reduce desktop hardware footprints by up to 4x.

o Time-sliced clusters: Like public cloud EC2 environments, Nutanix can provide a truly converged cloud infrastructure allowing you to run your server and desktop virtualization on a single converged cloud. Get the efficiency and savings you require with a converged cloud on a truly converged architecture.

13


| 13

4. Solution Design

With the Citrix XenDesktop on Nutanix solution you have the flexibility to start small with a single block and scale up incrementally a node, a block, or multiple blocks at a time. This provides the best of both worlds–the ability to start small and grow to massive scale without any impact to performance.

In the following section we cover the design decisions and rationale for the XenDesktop deployments on the Nutanix Complete Cluster.

Table 1: Solution Design Decisions

Item Detail Rationale

General

Minimum Size 1 x Nutanix block (4 ESXi hosts) Minimum size requirement

Scale Approach Incremental modular scale

Allow for growth from PoC (hundreds of desktops) to massive scale (thousands of desktops)

Scale Unit Node(s), Block(s), or Pod(s) Granular scale to precisely meet the capacity demands

Scale in n x node increments

VMware vSphere

Cluster Size Up to 8 ESXi hosts (Minimum of 4 hosts)

Isolated fault domains Citrix best practice

Clusters per vCenter Up to 2 x 8 host clusters Task parallelization

Datastore(s) 1 x Nutanix DFS datastore per pod (XenDesktop Server VMs, Provisioning Services Store, VM clones, VAAI clones, etc.)

Nutanix handles I/O distribution/localization

n-Controller model


Small deployments: Shared cluster

Large deployments: Dedicated cluster

Dedicated infrastructure cluster for larger deployments (best practice)

Nutanix

Cluster Size Up to 16 nodes Isolated fault domains

Storage Pool(s) 1 x Storage Pool (PCIe SSD, SATA SSD, SATA HDD)

Standard practice ILM handles tiering

Container(s) 1 x Container for VMs 1 x Container for Data (not

used here)

Standard practice

Citrix XenDesktop

XenDesktop Controllers

Min: 2 (n+1) Scale: 1 per additional pod

HA for XenDesktop Controllers

14


| 14

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

Load Balancing NetScaler or Load Balancer Ensures availability of controllers

Balances load between controllers & pods

Citrix Provisioning Services

PVS Servers Min: 2 (n+1) Scale: 1 per additional pod

HA for PVS servers

Users per PVS Server

Up to 1,000 streams PVS best practice

Load Balancing Provisioning Services Farm Ensures availability of PVS servers

Balances load between PVS servers & pods

vDisk Store Dedicated disk on Nutanix Standard practice

Write Cache On local hard drive Best practice

Citrix Storefront

Storefront Servers Min: 2 (n+1)

HA for Storefront servers

Load Balancing NetScaler or Load Balancer Ensures availability of Storefront servers

Balances load between Storefront servers & pods

Citrix NetScaler (If used)

NetScaler Servers Min: 2 (n+1)

HA for NetScaler servers

Users per NetScaler Server

See product data sheet Varies per model

Load Balancing NetScaler Cluster Ensures availability of NetScaler servers

Balances load between NetScaler servers & pods

15


| 15

A high-level snapshot of the Citrix XenDesktop on Nutanix Pod highlights can be seen below

Table 2: Pod Highlights

Item Qty

Control Pod

# of vCenter Server(s) 1

# of XenDesktop Controller(s) 2

# of XenDesktop Storefront(s) 2

Services Pod

# of Nutanix Blocks Up to 4

# of ESXi Hosts Up to 16

# of Nutanix Cluster(s) 1

# of Datastore(s) 1

Services Pod

8 Host Cluster8 Host Cluster

Nutanix Block Nutanix Block Nutanix Block Nutanix Block

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

Nutanix DFS (NDFS)

NFS Datastore

Streamed/Hosted Virtual Desktops

VMware vCenter

Control Pod

Citrix XenDesktop Controller


Load BalancerLoad Balancer


Figure 5 XenDesktop Pod Overview

The section below describes the desktop sizing and considerations for hosted virtual and streamed desktops.

16


| 16

Desktop Sizing

The following are examples of some typical scenarios for desktop deployment and utilization.

Table 3: Desktop Scenario Definition

Scenario Definition

Task Workers 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 by knowledge workers. Task workers who work specific shifts might all log in to their virtual desktops at the same time. Task workers include call center analysts, retail employees, and warehouse workers.

Knowledge Workers Knowledge workers’ daily tasks include accessing the Internet, using email, and creating complex documents, presentations, and spreadsheets. Knowledge workers include accountants, sales managers, and marketing research analysts.

Power Users Power users include application developers and people who use graphics-intensive applications.

Below are some initial recommendations for desktop sizing for a Windows 7 desktop. Note: These are recommendations for sizing and should be modified after a current state analysis.

Table 4: Desktop Scenario Sizing

Scenario vCPU Memory Disks

Task Workers 1 1GB 30GB (OS)

Knowledge Workers 1-2 2GB 30GB (OS)

Power Users 2 4GB 30GB+ (OS)

Desktop Optimizations

Following are some high-level desktop optimizations we followed for this design:

▫ Size desktops appropriately for each particular use case.

▫ Use a mix of applications installed in gold images and application virtualization, depending on the scenario.

▫ Disable unnecessary OS services and applications.

▫ Redirect home directories or use a profile management tool for user profiles and documents.

For more detail on desktop optimizations refer to the Citrix XenDesktop Windows 7 Optimization Guide document on http://support.citrix.com/

http://support.citrix.com/article/CTX127050

17


| 17

4.1. XenDesktop Machine Creation Services (MCS)

Citrix Machine Creation Services utilizes a standardized model for hosted virtual desktop creation. Leveraging a base, or “Master VM”, MCS will create clone VMs which consist of a delta and identity disk which link back to the base VMs disks.

In the figure below we cover the main architectural components of an MCS deployment on Nutanix and show the communication path between services.

StorefrontXenDesktop Controllers

Hosted Virtual DesktopsHosted Virtual Desktops

vCenter Server

Master VM(s)

Brokering

1 x NutanixNFS Datastore

Microsoft DFS(DFS-N/DFS-R)

MCS Ops

VM Ops

Personal vDisk

User DataDevice Access

Figure 6 MCS Communication

18


| 18

MCS Pod Design

A high-level snapshot of the Citrix XenDesktop on Nutanix Hosted Virtual Desktop Pod highlights can be seen below

Table 5: MCS Pod Detail

Item Qty

Control Pod




Services Pod




# of Datastore(s) 1

# of Desktops Up to 1200

Services Pod



ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

Nutanix DFS (NDFS)

NFS Datastore

Hosted Virtual DesktopsUp to 1200 Desktops

VMware vCenter

Control Pod





Figure 7 MCS Pod Detail

19


| 19

Hosted Virtual Desktop IO path with MCS

The figure below describes the high-level IO path for a MCS based desktop on Nutanix. As shown, all IO operations are handled by NDFS and occur on the local node to provide the highest possible IO performance. Read requests for the Master VM occur locally for desktops hosted on the same ESXi node and over 10GbE for desktops hosted on another node.

10

Gb

E1

0 G

bE

Desktop 1..N

OS OS DeltaPersonal

vDisk

ESXi Host

Nutanix CVM

ReadCache

SATASSD

SATAHDD

PCIeSSD

ILM

READ IOREAD/WRITE IO

ESXi Host

Nutanix CVM

ReadCache

SATASSD

SATAHDD

PCIeSSD

Master VM

OS

Desktop 1..N

OS OS DeltaPersonal

vDisk

ILM

Node 1Node 1Node NNode N

Nutanix DFS (NDFS)

Figure 8 MCS IO Overview

The figure below describes the detailed IO path for a MCS based desktop on Nutanix. All write IOs will occur locally on the local node’s PCIe SSD tier to provide the highest possible performance. Read requests for the Master VM occur locally for desktops hosted on the same ESXi node and over 10GbE for desktops hosted on another node. These reads are served from the high performance read cache (if cached) or the PCIe SSD tier. Each node will also cache frequently accessed data in the read cache for any local data (delta disks, personal vDisks (if used)). Nutanix ILM will continue to constantly monitor data and the IO patterns to choose the appropriate tier placement.

10

Gb

E1

0 G

bE

Desktop 1..N

OS OS DeltaPersonal

vDisk

ESXi Host

Nutanix CVM

ReadCache

SATASSD

SATAHDD

PCIeSSD

ILM


ESXi Host

Nutanix CVM

ReadCache

SATASSD

SATAHDD

PCIeSSD

Master VM

OS

Desktop 1..N

OS OS DeltaPersonal

vDisk

ILM


Figure 9 MCS IO Detail

20


| 20

4.2. XenDesktop Provisioning Services (PVS)

Citrix Provisioning Services streams desktops over the network from a centralized store of master vDisks (OS images). These vDisks are stored by the PVS server and are delivered by the Citrix Stream service. Upon boot the streamed desktop will pull the configuration via PXE/TFTP and will then initiate communication with the PVS server to continue booting the vDisk.

In the figure below we cover the main architectural components of a PVS deployment on Nutanix and show the communication path between services.

StorefrontXenDesktop Controllers

Streamed DesktopsStreamed Desktops

vCenter Server

Master vDisks

Stream

Brokering

DFS-R

vDisk StorevDisk Store

Provisioning Services (PVS)

1 x NutanixNFS Datastore

Microsoft DFS(DFS-N/DFS-R)

VM Ops

Catalog Ops

Target DeviceOps

Write Cache

Personal vDisk

Write Cache

Home Folder Redirection

Device Access

Figure 10 PVS Communication

21


| 21

PVS Pod Design

A high-level snapshot of the Citrix XenDesktop on Nutanix Streamed Desktop Pod highlights can be seen below

Table 6: PVS Pod Detail

Item Qty

Control Pod




# of PVS Server(s) 2

Services Pod




# of Datastore(s) 1

# of Desktops Up to 1200

Services Pod



ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

ESX

i

Nutanix DFS (NDFS)

NFS Datastore

Streamed DesktopsUp to 1200 Desktops

VMware vCenter

Citrix PVS ClusterCitrix PVS Cluster



Control Pod





Figure 11 PVS Pod Detail

22


| 22

PVS Store and Network Mapping

In the figure below we show the mapping for the PVS server’s storage and network. In this case we used dedicated interfaces for both PVS server management and Stream services.

Nutanix DFS (NDFS)

RF2 CTR VM

40GB VMDK

OS + App (C:)

PV-SCSI Controller

NFS Datastore

250GB+ VMDK

Store (S:)

SP1 (PCIe SSD,SATA SSD, SATA HDD)

Provisioning Services VM

Management PortGroup

PVS PXE PortGroup

vSwitch/vDS

Mgmt NIC Stream NIC

vmxnet3

Infra/AD

vmxnet3

PXE/TFTP/StreamServices

Figure 12 PVS Component Mapping

23


| 23

Streamed Desktop IO path with PVS

The figure below describes the high-level IO path for a streamed desktop on Nutanix. All write IO operations are handled by NDFS and occur on the local node to provide the highest possible IO performance. Streamed desktops hosted on the same server as the PVS host will be handled by the hosts local vSwitch and not use the external network.

10

Gb

E1

0 G

bE

Streamed Desktop 1..N

OSWrite Cache

PersonalvDisk

ESXi Host

Nutanix CVM

ReadCache

SATASSD

SATAHDD

PCIeSSD

ILM

ESXi Host

Nutanix CVM

ReadCache

SATASSD

SATAHDD

PCIeSSD

vDisk

vDisk

PVS Server

OS MemoryvDiskStore


OSWrite Cache

PersonalvDisk

Stream

ILM


...

Nutanix DFS (NDFS)


Figure 13 PVS IO Overview

The figure below describes the detailed IO path for a streamed desktop on Nutanix. All write IOs (write cache or personal vDisks (if used)) will occur locally on the local node’s PCIe SSD tier to provide the highest possible performance. The PVS server’s vDisk store is hosted on the local node’s PCIe SSD tier and will also be cached in-memory. All read requests from the streamed desktop will then be streamed either from the PVS server’s memory or its vDisk store which is hosted on NDFS. Each node will cache frequently accessed data in the read cache for any local data (write cache, personal vDisks). Nutanix ILM will continue to constantly monitor data and the IO patterns to choose the appropriate tier placement.

10

Gb

E1

0 G

bE


OSWrite Cache

PersonalvDisk

ESXi Host

Nutanix CVM

ReadCache

SATASSD

SATAHDD

PCIeSSD

ILM

ESXi Host

Nutanix CVM

ReadCache

SATASSD

SATAHDD

PCIeSSD

vDisk

vDisk

PVS Server

OS MemoryvDiskStore


OSWrite Cache

PersonalvDisk

Stream

ILM


...


Figure 14 PVS IO Detail

24


| 24

4.3. Nutanix – Compute/Storage

The Nutanix virtual computing platform provides an ideal combination of both high-performance compute with localized storage to meet any demand. True to this capability, this reference architecture contains zero reconfiguration of or customization to the Nutanix product to optimize for this use case.

Figure 8 shows a high-level example of the relationship between a Nutanix block, node, storage pool and container

Nutanix Block

Nutanix Node

PC

Ie S

SD

SATA

SSD

SATA

HD

D

Nutanix Node

PC

Ie S

SD

SATA

SSD

SATA

HD

D

Nutanix Node

PC

Ie S

SD

SATA

SSD

SATA

HD

D

Nutanix Node

PC

Ie S

SD

SATA

SSD

SATA

HD

D

Nutanix Block

Nutanix Node

PC

Ie S

SD

SATA

SSD

SATA

HD

D

Nutanix Node

PC

Ie S

SD

SATA

SSD

SATA

HD

D

Nutanix Node

PC

Ie S

SD

SATA

SSD

SATA

HD

D

Nutanix Node

PC

Ie S

SD

SATA

SSD

SATA

HD

D

Storage Pool – SP01

Container 1 -CTR-RF2-VM-01

...

Container N...

Figure 15 Nutanix Component Architecture

Below we show the Nutanix storage pool and container configuration.

Table 7: Nutanix Storage Configuration

Name Role Details

SP01 Main storage pool for all data PCI-e SSD, SATA,SSD, SATA-HDD

CTR-RF2-VM-01 Container for all VMs ESXi – Datastore

CTR-RF2-DATA-01 Container for all Data (Not used here) ESXi – Datastore

25


| 25

4.4. Network

Designed for true linear scaling, we leverage a Leaf Spine network architecture. A Leaf Spine architecture consists of two network tiers: an L2 Leaf and an L3 Spine based on 40GbE and non-blocking switches. This architecture maintains consistent performance without any throughput reduction due to a static maximum of three hops from any node in the network.

The figure below shows a design of a scale-out Leaf Spine network architecture which provides 20Gb active throughput from each node to its L2 Leaf and scalable 80Gb active throughput from each Leaf to Spine switch providing scale from 1 Nutanix block to thousands without any impact to available bandwidth

Figure 16 Leaf Spine Network Architecture

26


| 26

5. Solution Application

This section applies this pod based reference architecture to real world scenarios and outlines the sizing metrics and components.

NOTE: Detailed hardware configuration and product models can be found in the appendix.

5.1. Scenario: 300 Desktops

Table 8: Detailed Component Breakdown - 300 Desktops

Item Value Item Value

Components Infrastructure

# of Nutanix Desktop Pods 1 (partial) # of vCenter Servers 1

# of Nutanix Blocks 1 # of ESXi Hosts 4

# of RU (Nutanix) 2 # of vSphere Clusters 1

# of 10GbE Ports 8 # of Datastore(s) 1

# of 100/1000 Ports (IPMI) 4

# of L2 Leaf Switches 2

# of L3 Spine Switches 1

Figure 17 Rack Layout - 300 Desktops

27


| 27

5.2. Scenario: 600 Desktops

Table 9: Detailed Component Breakdown - 600 Desktops



# of Nutanix Desktop Pods 1 (partial) # of vCenter Servers 1




# of 100/1000 Ports (IPMI) 8



Figure 18 Rack Layout - 600 Desktops

28


| 28

5.3. Scenario: 1,200 Desktops

Table 10: Detailed Component Breakdown - 1,200 Desktops



# of Nutanix Desktop Pods 1 # of vCenter Servers 1




# of 100/1000 Ports (IPMI) 16



Figure 19 Rack Layout - 1,200 Desktops

29


| 29









# of 100/1000 Ports (IPMI) 32




30


| 30









# of 100/1000 Ports (IPMI) 64




31


| 31









# of 100/1000 Ports (IPMI) 128




32


| 32









# of 100/1000 Ports (IPMI) 256




33


| 33

6. Validation & Benchmarking

The solution and testing provided in this document was completed with Citrix XenDesktop 5.6 deployed on VMware vSphere 5.1 on Nutanix virtual computing platform.

The Login VSI Light and Medium benchmarks were leveraged to detail the desktop performance for a knowledge user on the Nutanix appliance.

6.1. Environment Overview

An existing Nutanix NX-2400 was used to host all infrastructure and XenDesktop services, as well as the Login VSI test harness. A Nutanix NX-3400 was utilized as the target environment and provided all desktop hosting. Both Nutanix blocks were connected to an Arista 7050S top-of-rack switch via 10GbE.

Infrastructure ServicesInfrastructure Services

Active Directory

vCenter DFS Server

vCOps

LoginVSI Test EnvironmentLoginVSI Test EnvironmentDesktopsDesktops

Sessions

LoginVSI Launcher(s)

Citrix XenDesktop ServicesCitrix XenDesktop Services

Provisioning Services

XenDesktop Controllers

Web Interfaceon NetScaler

Figure 24 Test Environment Overview

34


| 34

Test Environment Configuration

Assumptions:

o Knowledge worker use case

o Per-desktop IOPS (Task Worker): 5 sustained/70 peak (boot)

o Per-desktop IOPS (Knowledge Worker): 10 sustained/70 peak (boot)

o Using both MCS and PVS

Hardware:

o Storage/Compute: 1 Nutanix NX-2400, 1 Nutanix NX-3400

o Network: Arista 7050Q(L3 Spine)/7050S(L2 Leaf) Series Switches

Desktop Configuration:

o OS: Windows 7 SP1 x64

o 1 vCPU & 2 GB memory (locked)

o 1 x 30GB OS Disk

o Applications:

▫ Microsoft Office 2007

▫ Adobe Acrobat Reader 9.1

▫ Internet Explorer

▫ Flash/video

Login VSI:

o Login VSI 3.7 Professional

XenDesktop Configuration:

Table 15: XenDesktop Configuration

VM Qty vCPU Memory Disks

XenDesktop Controller(s) 2 4 8 1 x 40GB (OS)

PVS Server(s) 2 4 32 1 x 40GB (OS) 1 x 250GB (Store)

Storefront Server(s) 2 8 8 1 x 40GB (OS)

35


| 35

Test Image Preparation – MCS

1. Create Base VM

2. Install Windows 7

3. Install Standard software

4. Optimize Windows 7

5. Add Machine to Domain

6. Install Citrix VDA

7. Install Login VSI Components

8. Create Snapshot

9. Create Clones using XenDesktop Setup Wizard

Test Image Preparation – PVS

1. Create Base VM

2. Install Windows 7

3. Install Standard software

4. Optimize Windows 7

5. Install PVS Target Device

6. Create vDisk

7. Set Bios to boot from PXE

8. Remove VMDK

9. Boot VM from vDisk (Private Mode)

10. Add Machine to Domain

11. Install Citrix VDA

12. Install Login VSI Components

13. Create disk for write cache

14. Convert to template

36


| 36

15. Convert vDisk (Standard Mode)

16. Create Clones using XenDesktop Setup Wizard

Test Execution

Analyze results (Login VSI/vCops)

Wait for test execution to finish

Start test

Set test parameters and number of sessions

Login to VSI Management Console

Reboot/Boot Login VSI Launcher(s)

Reboot/Power-On Desktops

37


| 37

6.2. Login VSI Benchmark

Login Virtual Session Indexer (Login VSI) is the de-facto industry standard benchmarking tool to test the performance and scalability of centralized Windows desktop environments like Server Based Computing (SBC) and Virtual Desktop Infrastructures (VDI).

Login VSI is 100% vendor independent and is used to test virtual desktop environments like Citrix XenDesktop and XenApp, Microsoft VDI and Remote Desktop Services, VMware View or any other Windows based SBC or VDI solution.

Login VSI is used for testing and benchmarking by all major hardware and software vendors, and is recommended by both leading IT-analysts and the technical community. Login VSI is vendor independent and works with standardized user workloads, therefore conclusions that are based on Login VSI test data are objective, verifiable and replicable.

For more information about Login VSI visit http://www.loginvsi.com/

Login VSI Light Workflow

Below is a time base layout of the Login VSI Light workflow which runs on each desktop throughout test execution and simulates a task worker desktop:

Figure 25 Login VSI Light Workflow

http://www.loginvsi.com/

38


| 38

Login VSI Medium Workflow

Below is a time base layout of the Login VSI Medium workflow which runs on each desktop throughout test execution and simulates a knowledge worker desktop:

Figure 26 Login VSI Medium Workflow

39


| 39

6.3. 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 are the full time it takes for an application or task to complete (for example, launch Outlook) and is not in addition to traditional desktop response times. These do not refer to the round trip time (RTT) for network IO, rather the total time to perform an action on the desktop.

During the test all VMs are powered on and the workload is started on a new desktop every 30 seconds until all sessions and workload is active.

Evaluation is quantified 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 VSI index average is the average response time dropping the highest and lowest 2 percent

o VSImax: If reached, the maximum value of sessions launched before the VSI Index Average gets above the VSI Baseline x 125% + 3000ms

Based user experience and industry standards, we recommend ideal ranges for these values are kept below the following values:

Table 16: Login VSI Metric Values

Metric Value(ms) Rationale

Minimum Response <1,000 Acceptable ideal response time

Average Response <2,000 Acceptable average response time

Maximum Response <3,000 Acceptable peak response time

VSI Baseline <1,000 Acceptable ideal response time

VSI Index Average <2,000 Acceptable average response time

40


| 40

Login VSI Graphs

The Login VSI graphs show the values defined above over the launching of each desktop session. Below we show an example graph showing the test data. The y-axis is the response time in ms and the x-axis is the number of active sessions.

Performance Graphs

The performance graphs show a plot of the data as well as a trend line. Below we highlight the various aspects of the graphs:

41


| 41

The performance graphs show two main phases of the testing: the VM boot storm and the Login VSI test execution which performs the logins and then starts the workflow execution.

42


| 42

7. Results

7.1. MCS – 400 Light Desktops

Login VSI Light Results

During the testing with 400 desktops VSImax was not reached with a baseline of 925ms and average VSIindex of 993ms.

Cluster Metrics

At the peak of the test execution CPU utilization for the ESXi hosts peaked at 95.21% with an average utilization of ~53%. Memory utilization peaked at ~88.9%.

43


| 43

Nutanix Datastore Metrics

IOPS peaked at ~4,493 during the boot storm to refresh the desktops. Sustained IOPS were around ~2,004 during the test execution.

Command latency peaked at ~3.39ms for read and ~16.83ms for write ops during the test execution. Average op latency was ~1.38ms for read and ~7.2ms for write.

44


| 44

7.2. MCS – 300 Medium Desktops

Login VSI Medium Results


Cluster Metrics

At the peak of the test execution CPU utilization for the ESXi hosts peaked at 96.95% with an average utilization of ~49%. Memory utilization peaked at ~70.74%.

45


| 45



Command latency peaked at ~2.56ms for read and ~12.65ms for write ops during the test execution. Average op latency was ~1.45ms for read and ~7.49ms for write.

46


| 46

7.3. PVS – 400 Light Desktops

Login VSI Light Results


Cluster Metrics

At the peak of the test execution CPU utilization for the ESXi hosts peaked at 99.12% with an average utilization of ~63.9%. Memory utilization peaked at ~90.71%.

47


| 47

Provisioning Server Metrics

During the boot storm CPU utilization on the PVS server peaked at ~20.16% and decreased to an average steady state CPU utilization of ~4.9%.

Network utilization peaked during the boot process at ~1,098KBps RX and ~75,767KBps TX. After boot these values decreased to have an average of ~197KBps RX and ~19,408KBps TX.

48


| 48


IOPS peaked at ~2,846.6 during the boot storm to refresh the desktops. Sustained IOPS were around ~1,362 during the test execution.

During test execution over 99% of IOPS were write IOPS as most OS reads are streamed from the PVS server.

Command latency peaked at ~2.27ms for read and ~7.75ms for write ops during the test execution. Write latency had an average of ~3.6ms.

49


| 49

7.4. PVS – 300 Medium Desktops



Cluster Metrics

At the peak of the test execution CPU utilization for the ESXi hosts peaked at 98.51% with an average utilization of ~57.3%. Memory utilization peaked at ~70.97%.

50


| 50




51


| 51





52


| 52

7.5. PVS – 360 Medium Desktops



Cluster Metrics

At the peak of the test execution CPU utilization for the ESXi hosts peaked at 107.68% (Intel TurboBoost) with an average utilization of ~59.3%. Memory utilization peaked at 88.39%.

53


| 53




54


| 54





55


| 55

8. Further Research

As part of its continuous determination to deliver the best possible solutions, Nutanix will continue to research into the following areas:

o Performance optimizations.

o Scale testing.

o Detailed use-case application.

o XenApp configuration and testing.

o Personal vDisk configuration and testing.

o GPU offload/peripheral testing.

o Joint solutions with partners.

56


| 56

9. Conclusion

Desktop user density will be primarily driven by the available host CPU resources and not due to any IO or resource bottleneck for both MCS and PVS deployments on Nutanix. Login VSI Light test results showed densities of over 400 task worker desktops per 2U Nutanix appliance. However, most VDI deployments fit within the knowledge worker category which was validated at over 300 desktops per 2U appliance using the Login VSI Medium test. Sizing for the pods was determined after careful consideration of performance as well as accounting for additional resources for N+1 failover capabilities.

The MCS tests showed light IO footprints on the Nutanix platform with a peak of ~5,000 aggregate IOPS during the test runs, peaking during the boot storms. Sustained IOPS were light ranging from 2,000-3,000 during the test execution. IO Latencies averaged <2ms for read and <8ms for write during peak load.

The PVS tests showed light IO footprints on the Nutanix platform with a peak of ~2,600 aggregate IOPS during the test runs, peaking during the boot storms. Sustained IOPS were light ranging from 500-2,600 during the test execution. IO Latencies averaged <1ms for read and <4ms for write during peak load. PVS server CPU utilization peaked at ~20% utilization during the boot storm with an average steady state at ~5%. PVS server network TX peaked at ~75,000KBps (75MBps) during the boot storm with an average steady state of ~23,000KBps (23MBps).

The Citrix XenDesktop on Nutanix solution provides a single high-density platform for desktop and application delivery. This modular pod based approach enables these deployments to easily be scaled.

57


| 57

10. Appendix: Configuration

Hardware

o Storage / Compute

▫ Nutanix NX-2400

o Per node specs (4 nodes per 2U block):

CPU: 2x Intel Xeon X5660

Memory: 192 GB Memory

▫ Nutanix NX-3400

o Per node specs (4 nodes per 2U block):

CPU: 2x Intel Xeon E5-2660

Memory: 256 GB Memory

o Network

▫ Arista 7050Q - L3 Spine

▫ Arista 7050S - L2 Leaf

Software o Nutanix

▫ NOS 3.0.3

o XenDesktop

▫ 5.6 FP1

o Provisioning Services

▫ 6.1

o Desktop

▫ Windows 7 SP1 x64

o Infrastructure

▫ ESXi 5.1.0 patch 2

▫ vCenter 5.1.0 patch 2

VM o Desktop

▫ CPU: 1 vCPU

▫ Memory: 2 GB (locked)

58


| 58

o Storage:

▫ 1 x 30GB OS Disk on CTR-RF2-VM-01 NDFS backed NFS datastore

59


| 59

11. References

11.1. Table of Figures

Figure 1 Nutanix Architecture ..................................................................................................... 7

Figure 2 Machine Creation Services .......................................................................................... 9

Figure 3 Provisioning Services ..................................................................................................10

Figure 4 XenDesktop on Nutanix Conceptual Arch ...................................................................11

Figure 5 XenDesktop Pod Overview .........................................................................................15

Figure 6 MCS Communication ..................................................................................................17

Figure 7 MCS Pod Detail ..........................................................................................................18

Figure 8 MCS IO Overview .......................................................................................................19

Figure 9 MCS IO Detail .............................................................................................................19

Figure 10 PVS Communication .................................................................................................20

Figure 11 PVS Pod Detail .........................................................................................................21

Figure 12 PVS Component Mapping .........................................................................................22

Figure 13 PVS IO Overview ......................................................................................................23

Figure 14 PVS IO Detail ............................................................................................................23

Figure 15 Nutanix Component Architecture ...............................................................................24

Figure 16 Leaf Spine Network Architecture ...............................................................................25

Figure 17 Rack Layout - 300 Desktops .....................................................................................26

Figure 18 Rack Layout - 600 Desktops .....................................................................................27

Figure 19 Rack Layout - 1,200 Desktops ..................................................................................28




Figure 23 Rack Layout - 19,200 Desktops ................................................................................32

60


| 60

Figure 24 Test Environment Overview ......................................................................................33

Figure 25 Login VSI Light Workflow ..........................................................................................37

Figure 26 Login VSI Medium Workflow .....................................................................................38

11.2. Table of Tables

Table 1: Solution Design Decisions ...........................................................................................13

Table 2: Pod Highlights .............................................................................................................15

Table 3: Desktop Scenario Definition ........................................................................................16

Table 4: Desktop Scenario Sizing .............................................................................................16

Table 5: MCS Pod Detail ...........................................................................................................18

Table 6: PVS Pod Detail ...........................................................................................................21

Table 7: Nutanix Storage Configuration ....................................................................................24

Table 8: Detailed Component Breakdown - 300 Desktops ........................................................26

Table 9: Detailed Component Breakdown - 600 Desktops ........................................................27

Table 10: Detailed Component Breakdown - 1,200 Desktops ...................................................28




Table 14: Detailed Component Breakdown - 19,200 Desktops .................................................32

Table 15: XenDesktop Configuration .........................................................................................34

Table 16: Login VSI Metric Values ............................................................................................39

61


| 61

12. About the Author

Steven Poitras is a Solutions Architect on the Technical Marketing team at Nutanix, Inc. In this role, Steven helps design architectures combining applications with the Nutanix platform creating solutions helping solve critical business needs and requirements and disrupting the infrastructure space. Prior to joining Nutanix he was one of the key solution architects at the Accenture Technology Labs where he was focused on the Next Generation Infrastructure (NGI) and Next Generation Datacenter (NGDC) domains. In these spaces he has developed methodologies, reference architectures, and frameworks focusing on the design and transformation to agile, scalable, and cost-effective infrastructures which can be consumed in a service-oriented or cloud-like manner.

Follow Steven on Twitter at @StevenPoitras

About Nutanix

Nutanix is the recognized leader in the emerging Virtual Computing Platform market. The Nutanix solution converges compute and storage resources into a single appliance, delivering a powerful, modular building block for virtual datacenters. It incorporates the same advanced, distributed software architecture that powers leading IT innovators such as Google, Facebook and Amazon – but is tailored for mainstream enterprises and government agencies. The Nutanix solution enables easy deployment of any virtual workload, including large-scale virtual desktop initiatives (VDI), development/test apps, big data (Hadoop) projects and more. Nutanix customers can radically simplify and scale out their datacenter infrastructures with cost-effective appliances that can be deployed in under 30 minutes for rapid time to value.

Follow the Nutanix blogs at http://www.nutanix.com/blog/

Follow Nutanix on Twitter at @Nutanix

http://twitter.com/StevenPoitras

http://www.nutanix.com/

http://www.nutanix.com/blog/

http://twitter.com/nutanix

62


| 62

Documents

White Paper - Citrix XenDesktop on vSphere With Nutanix