VMware

This Question is Answered

7 Replies Last post: Nov 5, 2009 5:40 AM by regnak  

Network performance vs Local launch posted: Nov 3, 2009 10:10 AM

Click to view regnak's profile Enthusiast 32 posts since
Aug 18, 2008

Hi Folks,

We've Thinapp'd Office 2003 & Office 2007 for a client. After troubleshooting mdm.exe / debugger services causing issues on launch we've gotten the app to run in 3 seconds when launched from a physical windows XP PC or 8 seconds if launched from a Network File Server (Server 2003). We've tried 3 different file servers, both physical and virtual and can't seem to bring down the launch time when it's loaded over the network.

Task Manager shows a CPU hit in both cases for most of that launch time, Network activity is around 12% max.

Can anyone give me a benchmark launch time for streaming Thinapp'd Office 2003 or 2007 over a LAN (gigabit ethernet)? I'd like to see if we're getting a good result or we need to take things further with VMware?

Thanks!

Mike


Re: Network performance vs Local launch

2. Nov 4, 2009 12:39 AM in response to: regnak
Click to view pbjork's profile Hot Shot 105 posts since
Mar 7, 2008

Hi..
The kind of performance you are experience is very normal. Even quite good. Every new version of ThinApp we optimize more and more so this time will probably be lower if you rebuild with newer versions of ThinApp in the future but for now I would be very happy with an 8 second launch of an Office application.


You are correct that a warm second launch of the application will reuse resources and therefore launch faster. We are not caching stuff on the hard disk but in memory will there be things being kept by the OS. A 100% correct explanation to this "caching" needs to be provided by one of our developers.

Re: Network performance vs Local launch

4. Nov 4, 2009 11:34 AM in response to: regnak
Click to view Theike's profile Hot Shot 124 posts since
Jun 13, 2008
Hi,

In your 'warm start' experience you have the advantage of several layers of caching.

First of all there is the file access cache, which on a 'idle' system will be containing quite a lot of the recently accessed storage. Idle in this case is a normal stand-alone system, that did not have any other operations (applications) loaded between sessions. To eliminate this, you could start some other programs between your first and second test start.

Next on a second start of the application the 'route' to the server you want to start the application from will be established. You will be fully authenticated and the server knows you. Most likely the server also has the files accessed in his 'file cache'. To make the test similar you could access the server and copy the exe (not start it, copy it) so the server might have the app in cache as well (and you are fully authenticated). Second could be to disconnect from the server (simplest: restart server) before the second run.

At last it might still be that there are some background processes active that keep the application 'in memory'. mdm and ctfmon are the most common annoyances for this. try to delete the sandbox. if you fail: app is still active! In that case a 2nd start will be a lot faster.

Of course you have the same test conditions created... i assume... Initializing the sandbox costs some time... Delete it for both tests.

Do all of the above and you will see the times will become quite comparable.

Kind regards,
Michael Baars

Re: Network performance vs Local launch

5. Nov 4, 2009 1:01 PM in response to: regnak
Click to view pbjork's profile Hot Shot 105 posts since
Mar 7, 2008
Sorry but there is no really good paper on ThinApp execution performance because it depends on so much. The applications, how they are packaged, the network, storage speed and so on.. If I am not mistaken I think some launch times are mentioned in the deployment guide published on the ThinApp product page, under resources.

Re: Network performance vs Local launch

6. Nov 4, 2009 1:12 PM in response to: regnak
Click to view pbjork's profile Hot Shot 105 posts since
Mar 7, 2008

Just some thoughts about this scenario. So natively installed Office will launch faster than a ThinApped one. This is a fact. But what do you gain in ThinApping it? If you stream the package you will save disk space (very important in a VDI scenario). OS stability, having the applications ThinApped will leave your OS clean and therefore much more stable. Boot time.. Having no applications installed will make the OS start much faster. Run performance, we all know that a clean OS without a lot installed runs much faster than one with a lot applications installed.


I think it is important to see the big picture and not only execution time. If your machine is more stable (lesser reboots) and it starts much faster and runs much faster you might at the end of the day save time running the apps ThinApped.


If MS Office should be ThinApped or not is a great discussion with no right or wrong answer. Most of the times Office ends up on the image because so many applications use Office so having it packaged forces you to AppLink a lot.

VMware Developer

SDKs, APIs, Videos, Learn and much more in the Developer community.

Learn More

Developer Sample Code

Increase your developer productivity with VMware API sample code.

Learn More

VMworld Sessions & Labs

Online access to the latest VMworld Sessions & Labs and online services.

Learn more

Purchase PSO Credits Online

Purchase credits to redeem training and consulting services online.

Buy Now

Community Hardware Software

View reported configurations or report your own.

Learn More

VMware vSphere

Come witness the next giant leap in virtualization.

Register Today

Communities