Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

VMware WP WorkloadConsiderations WP en

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Workload Considerations for

Virtual Desktop Reference


Architectures
I N F O R M AT I O N G U I D E
Workload Considerations for
Virtual Desktop Reference Architectures

Table of Contents

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
RAWC Workload. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Active Directory Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
RAWC GUI. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Configuration Tab. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Test Variables and Considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Application Launch Order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Application Launch Delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Total Number of Virtual Machines Under Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Test Mode. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Logoff Users . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Workload Variables and Considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Number of emails +1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Typing Speed in Seconds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Number of Random Words. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
PDF Pages to Browse. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
About the Author. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Acknowledgements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
White Papers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

I N F O R M AT I O N G U I D E / 2
Workload Considerations for
Virtual Desktop Reference Architectures

Introduction
As interest in virtual desktop infrastructure continues to accelerate, so do the needs for improved assessment,
planning, and deployment methodologies and tools. The historical methodology for determining deployment
parameters, (that is the impact of VDI on network, storage, and compute resources) typically failed to truly
assess the impact of the workload that would eventually end up being generated by the desktop virtual
machine once it went into production. The reason for this gap was simple : There were few, if any, workloads
that accurately represented what a user actually did during their workday, that is, the applications they used,
the frequency of use, and the intensity of use. So professional services organizations typically estimated
network and storage impacts based on their experience, and then architected or modified an environment
based on those estimates. The problem, of course, is that such qualitative assessments can underestimate the
impact of a deployment and result in storage contention or network bottlenecks, or overestimate the impact
and result in excess costs.
The purpose of this paper is to introduce a desktop workload tool that generates a realistic, adjustable
workload with various applications in the desktop virtual machine. The results gathered (regarding CPU usage,
memory utilization, storage, and network) can then be analyzed to identify appropriateness / readiness of a
given environment to run virtual desktops. This tool, called the Desktop Reference Architecture Workload Code
(RAWC), has been used in numerous studies to simulate application workloads for various user types. It can be
configured to simulate light, medium, or heavy user characteristics, including the types of applications used in a
typical Windows desktop environment.
RAWC can be used to evaluate server and storage performance, validate configurations, and perform scalability
studies and proof of concepts.

Description
The RAWC workload runs on a Windows 7 or XP guest operating system and is executed on each desktop
virtual machine on one or more ESX hosts. The RAWC workload has a set of functions that performs operations
on common desktop applications including Microsoft Office, Adobe Reader, McAfee Virus Scan, Windows Media
Player, Java, and 7-Zip. The applications are called randomly and perform operations that mimic those of a
typical desktop user, including open, save, close, minimize and maximize windows, view an html page, insert
text, insert random words and numbers, conduct a slideshow, view a video, run a virus scan, send and receive
email, and compress files.
The RAWC workload uses a configuration file that is created via the RAWC GUI and writes application open/
close times and any errors to log files in a shared network folder. Various test variables can be configured
via the RAWC GUI, including a start delay for creating boot storms, and density (delay between application
operations) resulting in applications under test to be run together faster, number of emails created and sent,
and typing speed.
The combination of applications and test variables that are configured can either increase or decrease the
workload on virtual machines and VMware ESX server(s). The goal of doing this is to evaluate how many virtual
machines per solution can be run.
In addition to running native applications, RAWC can also run VMware ThinApp applications hosted on either
a local or remote site.
RAWC supports the use of Active Directory Groups. Large-scale deployment evaluations will find this feature
extremely valuable. Based on the virtual machines Active Directory Group membership, a workload can be
configured specifically for that Group via the RAWC GUI. This allows virtual machines to run varying workloads
to simulate a more realistic large-scale work environment.

I N F O R M AT I O N G U I D E / 3
Workload Considerations for
Virtual Desktop Reference Architectures

RAWC was recently updated to support a number of new features including duration or time-based testing,
running applications in static order for repeatability, and the ability to log the user out after a test is completed.

Architecture
The RAWC Architecture was designed with the following in mind:
Simplicity Minimal number of components and software packages to install.
Ease of use GUI used to configure workloads, create log folders and launch and cleanup configuration files.
Scalability Unlimited number of virtual machines under test.
Active Directory aware Able to determine Group membership and locate the correct configuration file
for the test.
Policy based workload Configure realistic workloads based on Group membership for large scale
enterprise testing.

Figure 1 - RAWC Architecture

The components of the RAWC Architecture include the following:


Session Launcher Virtual Machine(s): One or more session launcher virtual machines must be set up
to support the launching of desktop sessions. Each session launcher virtual machine can support up to
20 sessions.

I N F O R M AT I O N G U I D E / 4
Workload Considerations for
Virtual Desktop Reference Architectures

Target Desktop Virtual Machines Workload: The RAWC code resides on each of the desktop virtual machines.
Email Server Virtual Machine: The email server is required only if you are running Microsoft Outlook. You may
use the email server that is provided or supply your own.
RAWC Controller Virtual Machine: The RAWC Controller hosts the RAWC GUI and the shared network folder
for the configuration and log files. The RAWC Controller can be a physical or virtual machine.

RAWC Workload
Applications
The RAWC workload consists of a number of applications most commonly used in a Windows desktop
environment. These applications are called randomly and the operations are described below.
Microsoft Word (1 & 2) Open, minimize and maximize the window, insert text, follow the cursor, save
modifications, and close
Microsoft Word (Random) Open, minimize and maximize the window, write random words and numbers,
save modifications, and close
Microsoft Excel Open, minimize and maximize the window, write random numbers in cells, insert and delete
a column and row, copy and paste formulas, save modifications, and close
Microsoft PowerPoint Open a presentation, minimize and maximize the window, conduct a slide show
presentation, and close
Microsoft Outlook Setup of a user account
Microsoft Outlook Open, send an email, minimize and maximize the window, send additional emails, execute
a send/receive, and close
Internet Explorer Open, browse an html page, minimize and maximize the window, and close
Windows Media Player Open, view a video, and close
Adobe Acrobat Reader Open a PDF document, browse random pages, minimize and maximize the window,
and close
Java Open, run a java application, and close
McAfee Anti-virus Execute a real time virus scan
7-Zip Open, compress large files, and close

Active Directory Groups


The RAWC workload, if configured for Active Directory Groups, will query the Active Directory for the virtual
machines Group membership, then locate the proper configuration file and begin the test. This feature allows
workloads to be defined based on the virtual machines Group membership, thus enabling multiple workloads
to run in the same test. For large-scale tests, this allows for a more realistic representation of an enterprise. For
example: All groups need access to email, Internet Explorer and Adobe Reader, but Finance needs access to
Excel and PowerPoint, whereas an individual in Manufacturing may not.

I N F O R M AT I O N G U I D E / 5
Workload Considerations for
Virtual Desktop Reference Architectures

RAWC GUI
The RAWC GUI is used to configure RAWC. The GUI enables you to save and retrieve test variables, create
log folders, and define unique workloads based on Active Directory Groups. In addition, you can use the GUI
to increase or decrease the load or adjust the user behavior, such as the number of words per minute. The
Configuration tab contains information about the test, logging information, email variables, and applications
selected for each Group, if using Active Directory Groups.
Configuration Tab
The Configuration tab contains two sections, General and Workload.

Figure 2 - Configuration Tab

Test Variables and Considerations


The General section allows you to configure the test variables.
Application Launch Order
This feature allows you to determine if applications will run in a random or static order.
Random Order
Random order is the standard mode in RAWC. By running applications in a random order, applications will be
executed in a different (random) order on different virtual machines. When running applications in a random
order, it is important to understand the following:
Applications run in a random order.
A varying Start Delay (explained in the next section) is in effect, depending on the number selected for the
Total number of virtual machines under test.
The random delay between applications (Application Launch Delay) is calculated on the fly. When running
applications in random mode, the user chooses the upper limit number used in the random delay calculation.

I N F O R M AT I O N G U I D E / 6
Workload Considerations for
Virtual Desktop Reference Architectures

Static Order
Static order is a new feature in RAWC. By running applications in a static order, all virtual machines will run
the exact same test, including the applications running in the same order and with fixed delays between the
applications. This feature gives you the ability to repeat a test over and over again. When running applications
in a static order, it is important to understand the following:
The applications selected will run in the exact same order for each iteration on each virtual machine.
There is no Start Delay (explained below) in effect. A varying Start Delay in this situation is irrelevant, as the
exact same applications are running on each virtual machine, so if the test started in one minute or 12 minutes
or 40 minutes, all virtual machines would start together.
The random delay between applications (Application Launch Delay) is now a fixed delay. When running
applications in static order, the upper limit number that is selected becomes the actual delay between the
applications.
Application Launch Delay
The application launch delay is used to either increase or decrease the load on the server by running the
applications closer or farther apart. Keep the following in mind:
When running applications in a random order, the number selected for the application launch delay becomes
the upper limit of a random delay that is calculated on the fly. For example, the random delay that is
calculated could be 1 second or 119 seconds or any number in between if 120 seconds is selected.
When running applications in a static order, the number selected for the launch delay, becomes the actual
fixed delay between applications.
Typically, the greater the number of virtual machines under test, the larger a number you will want to select,
say above 120 seconds.
By selecting a lower number, you can generate a heavier load by making the applications run closer together.
Total Number of Virtual Machines Under Test
The total number of virtual machines under test is used to determine the Start Delay of the virtual machine.
This feature allows you to simulate a normal login scenario or create a more intensive Boot Storm. A Boot
Storm occurs when more virtual machines start their workloads closer together, thus creating a larger load
on the server. A Boot Storm scenario is likely to occur after systems have recovered from a power outage or
logging in first thing in the morning.
A normal login scenario allows virtual machines to start their workloads over a period of time. Select the
number of virtual machines that is closest to the number of virtual machines you have under test.
- 10 virtual machines under test No start delay
- 64 virtual machines under test 0 to 12 minutes
- 250 virtual machines under test 0 to 24 minutes
- 500-1000 virtual machines under test 0 to 54 minutes
A Boot Storm scenario can be created by simply selecting a lesser number of virtual machines under test.
For example: If you are running a 500 virtual machine test, select 250 for the Total number of virtual
machines under test. The means that 500 virtual machines will start within 0 to 24 minutes of each other.

I N F O R M AT I O N G U I D E / 7
Workload Considerations for
Virtual Desktop Reference Architectures

Test Mode
The GUI allows you to select either the number of iterations or the duration (minutes) that a RAWC test will run.
Iterations
The GUI allows you to select the number of iterations you want the virtual machines to run. If you are interested
in the open and close times of applications, you will want to run at least three iterations. Applications take
longer to open the first time. By the second iteration, applications/pages have been cached and the open times
are much shorter.
Duration
Duration is a new feature in RAWC. The GUI allows you to select the number of minutes you want the test to
run. You no longer have to guess at how many iterations it will take to run a four hour or even eight hour test.
When using duration, it is important to understand the following:
Virtual machines do not finish at exactly the same time or exactly on the duration time.
Time checks are performed before and after random (or static) delays between applications.
Time checks are performed at certain points while an application is performing. For example:
- A time check occurs before and after an email is sent, but not during the creation of an email.
- A time check occurs before and after the slide presentation.
When the duration time has been met, the log files are closed and the desktop is cleaned of any
open applications.
If you are using Active Directory Groups and running in the duration test mode, you will notice that
lighter workloads (2-3 applications) will run more iterations of the applications than heavier workloads
(5+ applications). This is fairly accurate, as task workers are limited in the number of applications they run,
and more often than not, use these applications more often.
Logoff Users
Logoff users is a new feature in RAWC. You can now specify via the GUI if the user logs out of the virtual
machine after the test completes or remains logged in.

Figure 3 - Workload Section of Configuration Tab

I N F O R M AT I O N G U I D E / 8
Workload Considerations for
Virtual Desktop Reference Architectures

Workload Variables and Considerations


The Workload section allows you to define a unique workload and workload variables per Active Directory
Group. If your environment does not include an Active Directory, a group name of NA can be used. The
variables in this section are more reflective of user behavior, but can also be used to create an additional load
on the server.
Number of emails +1
You can configure the number of emails that are created and sent, via the GUI. Outlook is always the first
application that is opened, with an email being created and sent. Other applications are then opened randomly
and operations performed. Outlook is then randomly called later in the test and the number of emails specified
in the GUI are created and sent.
Typing Speed in Seconds
You can manipulate user behavior by selecting either a fast, medium, or slow typing speed. By selecting the fast
typing speed, you can move through the word applications quickly.
Number of Random Words
The random word document allows you to define the number of random words and numbers that are written to
the document. This particular application is able to create an additional load on the virtual machine since there
is a calculation to determine what word or number is written to the document.
PDF Pages to Browse
You can define the number of random pages that are browsed in the PDF document. To increase the load,
decrease the number of pages browsed so that the next application is accessed faster.
Results
The esxtop tool was used to monitor in real time and gather data about the state of the physical server running
an ESX Server. The esxtop tool is invoked from the command line and provides information on the CPU
utilization of each physical processor, memory utilization, and disk and network bandwidth for each disk and
network device available to the ESX Server machine.
Information about interpreting esxtop statistics can be found here:
http://communities.vmware.com/docs/DOC-9279
Application open and close times can be gathered from the timing logs deposited by each virtual machine
in the shared network folder on the Test Master virtual machine. This data can be manipulated and graphed
in Excel.

Conclusion
As the desktop virtualization world evolves, so will the desktop workload tools. Know your environment
(network, servers, storage and desktops), and your users (applications, light/heavy use), before determining
which workload tool to use. Research the various tools and even conduct an assessment to determine if
virtualization is right for your company. Make sure the tool is easy to install and use and that it simulates the
applications you run most. Be sure to allow enough time for a proper assessment, testing and interpretation of
the results, as this will aid in the proper planning and deployment of your virtual desktop infrastructure.

I N F O R M AT I O N G U I D E / 9
Workload Considerations for
Virtual Desktop Reference Architectures

About the Author


Fred Schimscheimer is a Senior Technical Marketing Manager at VMware. In this role, he works as part of the
product marketing team as an expert in storage and workloads for virtual desktop solutions. Fred is the architect
and developer of the RAWC workload.

Acknowledgements
I would like to thank Mason Uyeda for his continued support and encouragement in developing the RAWC
workload. I would like to give special thanks to Srinivasan Muthu for his expertise in designing and coding
the RAWC GUI and packaging. I would also like to thank Scott Drummonds for introducing me to workload
automation and providing the initial ideas for the workload.

White Papers
The following white papers used the RAWC workload:
Vblock Powered Solutions for VMware View 4.5 Reference Architecture
http://www.vce.com/pdf/solutions/vce-vmware-view-reference-architecture.pdf
VMware View 4.5 on Cisco Unified Computing System and EMC Unified Storage Design Guide
http://www.cisco.com/en/US/docs/solutions/Enterprise/Data_Center/Virtualization/ucs_view_emc.html
VMware View 4.5 on Cisco Unified Computing System and NetApp Storage Design Guide
http://www.cisco.com/en/US/docs/solutions/Enterprise/Data_Center/Virtualization/ucs_view_netapp.html
Simplifying enterprise desktop deployment and management using Dell EqualLogic storage and VMware View
http://www.vmware.com/files/pdf/Dell-VMware-View-RA.pdf
Desktop Sizing Guide: VMware View 4.0 and VMware vSphere 4.0 Update 1
http://www.mainline.com/_web/pdfs/mrktgMaterials/VMware-WP-VMsizing-050310.pdf
50,000-Seat VMware View Deployment
http://www.vmware.com/files/pdf/VMware-View-50kSeatDeployment-WP-EN.pdf
VMware View Reference Architecture A Guide to Large-scale Enterprise VMware View Deployments
http://www.vmware.com/files/pdf/resources/vmware-view-reference-architecture.pdf
VMware ThinApp Reference Architecture A Guide for Enterprise VMware ThinApp Deployments
http://www.vmware.com/files/pdf/thinapp-ref-arch.pdf
EMC CLARiiON Storage A Guide to deploying EMC CLARiiON CX4-240 FC with VMware View
http://www.vmware.com/files/pdf/view-deployment-emc-clariion-fc.pdf
EMC CLARiiON Storage A Guide to deploying EMC CLARiiON CX4-240 iSCSI with VMware View
http://www.vmware.com/files/pdf/view-deployment-emc-clariion-iscsi.pdf
EMC Celerra Unified Storage A Guide to deploying EMC Celerra NS20 storage with VMware View
http://www.vmware.com/files/pdf/resources/vmware-view-ns20-deployment-guide.pdf
Scalability Study for Deploying VMware View on Cisco UCS and EMC Symmetrix V-Max Systems
http://www.cisco.com/en/US/docs/solutions/Enterprise/Data_Center/App_Networking/vdiucswp.html
Using Local Disks in a VMware View Deployment
http://www.vmware.com/files/pdf/view_local_disk.pdf

I N F O R M AT I O N G U I D E / 1 0
Workload Considerations for
Virtual Desktop Reference Architectures

The VMware Reference Architecture for Cisco UCS and EMC CLARiiON with VMware View 4
http://vmware.com/go/vce-ra-brief
Deploying Virtual Desktops with VMware View 4 on EMC Celerra NS-120 Validation Test Report
http://www.vmware.com/files/pdf/VMware-View4-EMC-NS-120-VTR.pdf
VMware View on NetApp Storage Solution Brief
http://www.vmware.com/files/pdf/VMware-VMwareView-NetAppStorage-WP-EN.pdf

VMware, Inc. 3401 Hillview Avenue Palo Alto CA 94304 USA Tel 877-486-9273 Fax 650-427-5001 www.vmware.com
Copyright 2010 VMware, Inc. All rights reserved. This product is protected by U.S. and international copyright and intellectual property laws. VMware products are covered by one or more patents listed at
http://www.vmware.com/go/patents. VMware is a registered trademark or trademark of VMware, Inc. in the United States and/or other jurisdictions. All other marks and names mentioned herein may be
trademarks of their respective companies. Item No: VMW_10Q1_WP_WorkloadConsiderations_EN_P11_R2

You might also like