Checkouts visibly slow

asills · Post by **asills** » Thu Feb 24, 2005 10:56 am

When performing a checkout and subsequently a checkin of a single file without much history (one of our assemblyinfo.cs files for instance), the checkout can take as long as 12 seconds and the checkin (revert) can take 5 or 6 seconds (GUI client). This happens both from a client machine as well as directly on the server.

Using Visual Studio integration this action can take near 20 seconds while the VS.NET UI just hangs, then goes to a busy icon, then hangs and eventually returns with a file showing checked out.

Server specs (both SQL and IIS):
Windows Server 2003
2 Xeon 2.4Ghz Processors
1.5GB RAM
"Normal" HDD (2 striped drives)

Repository:
Revision 116000
3400 Folders
62000 Files
~5.4GB .mdf file (~3.4GB when backed up and compacted)

If you connect to a different repository with less history, the checkout time decreases twofold, suggesting to me that large repositories are "slow".

I'm not exactly sure what more information to give, or where exactly I should be taking this question...

Post by **jclausius** » Thu Feb 24, 2005 11:30 am

Is it safe to asume you are using Folder security?

If so, there were *some* optimizations made in the upcoming Vault 3.0.3 - which should be released very, very shortly.

asills · Post by **asills** » Thu Feb 24, 2005 11:58 am

You assume correctly. And on the repository that's faster, there is no folder security.

If I temporarily disable my folder security, does it delete the assignments I've already made?

Post by **jclausius** » Thu Feb 24, 2005 12:01 pm

No. Your assignments will remain in place.

Can I ask an related question? Do you currently have any users where the default or folder security is set to 0 ( ie. no read, no checkin, no add )? Or do all your users have some right ( although it may be limited to read, or read/checkin )?

asills · Post by **asills** » Thu Feb 24, 2005 12:04 pm

I have both actually. Some groups have no access, some have read access and some have full access (and all vary by repository location).

Post by **jclausius** » Thu Feb 24, 2005 12:08 pm

OK. Thanks for the info.

Vault 3.0.3 will be significantly faster than Vault 3.0.1/3.0.2 for the cases where a user has at least limited access to the repository's tree.

Vault 3.0.3 will be slightly faster than Vault 3.0.1/3.0.2 for the cases where a user has been denied access to any part of the repository's tree. This area is a bit more difficult as each user's rights may vary. We will be investigating ways to improve performance for this case in a future version of Vault.

asills · Post by **asills** » Thu Feb 24, 2005 12:32 pm

Wow, I disabled folder security and it literally takes no time to do anything. In your estimation, will the 3.0.3 speed improvements be similar to the speed without folder security?

Also, when it comes out can I safely upgrade the server and let the clients stay on 3.0.1?

Post by **jclausius** » Thu Feb 24, 2005 1:16 pm

asills wrote:Wow, I disabled folder security and it literally takes no time to do anything. In your estimation, will the 3.0.3 speed improvements be similar to the speed without folder security?

Not quite. With security enabled, the server still has to check a user's security assignments.

asills wrote:Also, when it comes out can I safely upgrade the server and let the clients stay on 3.0.1?

Yes. Within any major version, we strive to keep the protocols from changing. Note, there have been some nice improvements made on the client as well. People may want to upgrade as soon as they can, but a server upgrade does not force the issue.

asills · Post by **asills** » Thu Feb 24, 2005 6:33 pm

Speaking of upgrading, do you have any sort of notification list you send updates to when a new version is released?

Post by **jclausius** » Fri Feb 25, 2005 9:53 am

We publish release information in the Announcements forum - http://support.sourcegear.com/viewforum.php?f=15

If you subscribe to the Announcements RSS feed - http://support.sourcegear.com/rss.php?f=15, you would receive any release information through there.

HTH

asills · Post by **asills** » Tue Mar 01, 2005 9:48 am

Thanks, 3.0.3 seriously improved performance with folder security. I can't tell the difference between having it enabled and having it disabled.

Post by **jclausius** » Tue Mar 01, 2005 10:20 am

Also note, there was a Vault 3.0.4 release. I mention this because the changes might affect your setup, since you're running with folder security.

asills · Post by **asills** » Tue Mar 01, 2005 10:22 am

Yeah, no reason for me to get 3.0.4 based on what the release notes say.