Team Foundation Dogfood stats
I know John has been doing this for a while on his blog (http://blogs.msdn.com/johnlawr) but things have been so busy lately that he hasn't had time to post the latest numbers so I thought I'd save him the time and do it.
Below is the data from a report we ran on 10/3/05. A couple of things to note about it...
- # of recent users has been trailing off a bit as VS 2005 winds down for it's approaching release/launch and there are fewer people actively working on it every day. We are adding some new teams here soon in order to keep the overall load on the server on an increasing slope.
- Load (particularly version control) varies a lot from report to report based on when new branches are created and whether or not major integrations between branches have happened that cause lots of new download activity
- We installed proxies in North Carolina, North Dakota and India after the last dogfood refresh and that lowers the download volume (and load) somewhat
We are currently in the process of running an experiment to put all of the data (bugs and work items) for the entire division into a TFS installation and run load tests against it using VSTS load testing tools. This will result in a server in about 10X the data that our dogfood server currently has in it. I don't know yet how much more load it will be - we're collecting stats on divisional load now. When we've done it (probably gonna be a couple of months), I'll post something about how it went.
Users
Work items
-
Work items: 49,807
-
Areas and Iterations: 2,127
-
Work item versions: 416,722
-
Attached files: 12,825
-
Queries: 4,640
Version control
-
Files/Folders: 622,832/72,155
-
LocalVersion: 24.4 M
-
Total compressed file sizes: 59.9G (there's about a 3-4 to 1 compression between compressed size and uncompressed size)
-
Workspaces: 995
-
Shelvesets: 2,010
-
Checkins: 10,418
-
Pending changes: 27,429
Commands (last 7 days)
We've got a case study of our dogfood server on http://www.microsoft.com/downloads/details.aspx?familyid=43e098e5-9087-44d2-846f-d33ab3dd966a&displaylang=en It's a pretty good start but I'm looking to beef it up with a lot more detailed info so you can gauge what it takes to build/manage a system of this scale (it's actually not very hard :)) The load we put on the server(s) to support what you see above is actually not very high. We're currently only running about 10-15% server utilization. When I beef up the case study I'll include more detail on that kind of analysis.
Brian
