2007-09-28

Thanks to Kirk Alexander for sharing his notes.

Notes from production conference call this morning
Stanford Performance issues
Most important was streamlining the login : make queries term specific
Modifications to the group providers
Added indexes to meetings table and it dropped out of hotlist of queries
Stanford: nobody has programming access to backend system
No separate staging area at Berkeley
Daily merge/feed from SIS into sakai schema at 7:30 a.m.
All assignments are fixed in the post 2.4 branch
Stanford not using the assignments tool
Lance has learned that knowing when to use the profiler can
Save a ton of time
Stanford been plagued with MyWorkspace corruption. This
has prevented users from logging in. Last week's velocity template
errors were related to this. Resolved by dropping the myworkspace
Kevin Chen at Berkeley
Last Friday did emergency push to solve performance problems
They hit assignments trouble. They moved to post 2.4 branch and see a lot of performance
But they had to hack it because they are not using the post 2.4 GB
Lance: new issue with assignments and GB integration
Also seeing a number of out of Memory errors - this seems to be increasing in frequency
Profiler helps you find where it's being slow within seconds
They use YourKit Version 6/7
Captures all of the JDBC calls
Shows not only the slowest running sql but takes you back to the stacktrace including the bound parameters
Thomas says this is similar to JProfiler
MyWorkspace issues are related to new users...
Lance:
They are in the process (if we do have find duplicate assignment submissions) of making a tool for the consultants to work with the faculty to resolve the ones they cannot find programmatically. Tool would allow the faculty member to do it on their own.

Lance reported on new news that will be announced publicly very shortly.
IU is re-writing the assignments tool (V2) They will write it in RSF and not use Hibernate (just Spring JDBC)
Assignments resolution tool is RSF as well. In the process they would remove all the XML and normalize that data.
Will work further on integration with Gradebook. Might loose some functionality in GB like the check mark grade scale.