TFS performance

Dec 9, 2010 at 10:19 PM

We noticed that installing packages into projects that are connected to a TFS server is painfully slow.  In some cases we notice that a package install can have up to 10 minutes added to the install when the project is connected to our corporate TFS server.  Has anyone else noticed this?

 

I am thinking of trying to profile the process and see if I can find a profiler that can filter by namespace.  Any other thoughts on this?

Developer
Dec 9, 2010 at 10:23 PM

The tfs integration hasn't been touched in a while and was never that performant. It'd be great to get some numbers but we haven't paid alot of attention to this scenario. A good place to start would be the NuGet.TeamFoundationServer project.

Dec 11, 2010 at 2:42 AM

I spent some time this morning running the 1.0 branch code and performance was much better.  I then had a few other people try out the RC today and it seems like the performance problems with TFS were in the CTP 2 release.  The RC seems to have much better performance now.

Just to give you some data.

the CTP 2 took 7 minutes to install 12 packages in a chain.  It tool 45 minutes to uninstall them.

using the RC release the install took about a minute and the uninstall took about a minute.

Having this much of a difference, I think I am going to create a package to verify the version of the client console using the $host.version and force everyone to upgrade in order to install the new packages.

 

Dec 11, 2010 at 4:13 AM

Thanks for getting numbers!  Though frankly, the difference is a bit puzzling, as I'm not sure that logic has changed much.  Is there any chance that TFS was having a slow moment at that time?

Dec 11, 2010 at 4:20 AM
I have consistently seen those number through out the week and the slow test numbers we pulled today were followed by a very quick update and rerun the test with the RC.
Now that I have my packages all converted to the RC, I will be able to perform some side by side tests to help eliminate the possibility of server issues causing the problems.
On Fri, Dec 10, 2010 at 10:13 PM, davidebbo <notifications@codeplex.com> wrote:

From: davidebbo

Thanks for getting numbers! Though frankly, the difference is a bit puzzling, as I'm not sure that logic has changed much. Is there any chance that TFS was having a slow moment at that time?

To add a post to this discussion, reply to this email (nuget@discussions.codeplex.com)

To start a new discussion for this project, email nuget@discussions.codeplex.com

You are receiving this email because you subscribed to this discussion on CodePlex. You can unsubscribe or change your settings on codePlex.com.

Please note: Images and attachments will be removed from emails. Any posts to this discussion will also be available online at codeplex.com


Dec 11, 2010 at 4:57 AM

Sounds good.  It's entirely possible that some seemingly small change to the code is causing this change.  Note that TFS is all http based, so it might be interesting to run fiddler during the test to see if there is a drastic difference in the number of requests happening.