Kirx' Blog - kirxblog.wordpress.com

ProcjectVRC Whitepaper: Impact of Application Virtualization on VDI Performance | June 17, 2011


It has been a question “ever since”:

What is the perfomance impact of Application Virtualization?

Now, Project Virtual Reality Check (http://www.projectvrc.com) has released a Whitepaper that provides an idea of that. ProjectVRC compared Microsoft App-V, VMware ThinApp and Citrix XenApp Streaming running in Virtual Windows 7 machines.

While the impact is significant, there are some things to note…

Announcement

Read the official announcement on BrianMadden.com:

http://www.brianmadden.com/blogs/rubenspruijt/archive/2011/06/15/the-impact-of-application-virtualization-solutions-in-vdi-the-reality-check.aspx (short: http://bit.ly/ii0pCY)

Measurement Methodology

The way ProjectVRC measures is somewhat different from most classical “Performance Test” tools. Most of these tools process loops of well-defined tasks. The results then directly reflect the execution speed (the faster the tasks are processed, the better the result).

ProjectVRC (actually a joint venture of PQR and “my” Login Consultants) uses an index that reflects the maximum amount of simultaneous user sessions on a given piece of hardware.

Essentally, more and more user sessions are created on that hardware, and when the user sessions start to get slow or inresponsive, the number of users is counted.

For the Application Virtualization test the decision was made to measure against a VDI environment. (how many users, each having an own VM, can work with an acceptable response time on the hardware). Another approach could have been: “How many Remote Desktop Services / Terminal Server User Sessions can be hosted”, but this was not tested.

For the AppVirt impact tests, ProjectVRC took “localy/classicaly” installed applications as a baseline (or 100%). Values below 100 mean that – with a given Application Virtualization solution – less users can work on the same “iron” with an acceptable performance (measured by the application’s response times).

Overall Results

Well.. no. I won’t disclose any details here. Check the Whitepaper, it’s free. Anyway there are some things that I want to share with you:

Max User Sessions Impact

As you can see, each Application Virtualization solutions only allows a reduced number of users compared to VMs with locally installed applications. The result shows that – using a given deployment scenario – there are at least 20% less sessions possible by having the core business applications virtualized. Microsoft Application Virtualization, Citrix Streaming and VMware Thinapp have different impacts, while two of them are quite close together and the third one seems to have some difficulties. The grey bar at the top marks 100% for locally installed applications.

This is an example for a “streaming” scenario with a fully virtualized Office suite. If the Virtual Applications are pre-cached on the Clients or some of the Office applications are installed locally, the results are different.

The ProjectVRC Whitepaper includes quite detailed results from all AppVirt solutions, like the difference between pre-cached and streamed delivery, the usage of App-V’s Share Cache or differences in the transfer protocol. It includes various charts like the one above.

Results for Selected Resources

Beyond that overall results, ProjectVRC also investigated the impact on Memory or the application launch time.

It has put an focus onto Disk I/O, because esp. in VDI environments Disk I/O (that is Storage I/O) is a sginifcant cist drive and perfromance bottleneck.

It shows that Application Virtualization can decrease the Read I/Os, but probably will increase Write I/Os from the VM’s (or VM Host’s) perspective.

Conclusion

I think the answer is clear: Application Virtualization solutions, namely App-V, XenApp Streaming and ThinApp do have a significant impact on the amount of users that can work “acceptably” fast on a given hardware. While introducing AppVirt to VDI environments, Storage I/O requires additional attention.

Considerations

While the key message is clear, keep in mind:

  • The results are not 1:1 applicable for your environment. Project VRC:
    • Run all core applications, including MS Office, in a virtualized manner. This is not a Best Practice
    • Used a very well-defined workload/taskset that may/may not reflect the majority of your users
    • Evaluated VDI VMs, that are different from an RDS/TS/SBC environment. On Terminal Serves, you may seem significant different result.
  • The applications were not specially “tuned” for Application Virtualization. It is sometimes recommended to avoid installation of certain features
  • Your environment can be faster or slower with different resources that impacts the performance.

Unluckwise, ProjectVRC did not (yet?)

  • Compare App-V applications that were installed to C: instead of Q: (VFS vs. MNT install) (Applications have been sequenced to Q: what is considered to be faster)
  • Compare the impact on SBC hosts
  • Check the impact on the “network”, namely the required bandwidth and time to download the packages, the impact on CPU and RAM of the streaming server…

But hey, there is always room for improvements, right?

Download

You can download the Whitepaper for free directly from the Project VRC WebSite. Go to http://www.projectvrc.com and look for the “Phase IV” whitepaper.

Or check the announcement at http://bit.ly/ii0pCY

Advertisements

Posted in App-V, Client
Tags: ,

Leave a Comment »

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: