Forum .LRN Q&A: Performance impact of new dotlrn mount for every instance of a class

Collapse
1: Performance impact of new dotlrn mount for every instance of a class
Posted by Peter Alberer on 08/18/03 11:26 AM
Looking at the dotlrn-code i ask myself, why is there a new dotlrn package instance created for every instance of a class? I understand that this is the openacs way to do it, but will this not become a performance problem rather soon? All of those dotlrn packages and sub-packages have sitemap entries and the performance of the request-processor will degrade. If i think of all the classes and class-instances that will accumulute over the months i think this could become a problem. Supporting 500 class-instances with dotlrn (with an archive lasting 2 years) probably means a sitemap of 16.000 entries (500 classes with 8 sub-packages per class over 4 terms) just for the dotlrn-classes. What do others think of that?
Collapse
2: Re: Performance impact of new dotlrn mount for every instance of a class (response to 1)
Posted by Malte Sussdorff on 08/18/03 11:54 AM
Hi Peter, we use dotLRN for www.aiesec.net with 100k users in roughly 3000 communities (same behaviour as a class if I'm not utterly mistaken). And it runs reasonably well. We are at the moment finding the culprits and will provide some tuning suggestions once we are done with it. As for the sitemap, no, not a big issue. It takes 40 seconds to load all the entries into the cache, but this only happens during the startup, so it is fine.

To be fair, let me state the hardware:

- 2*933Ghz, 1GB RAM server with a seperate RAID system for the content-repository-files for the aolserver

- 2*Opteron242, 4GB RAM with multi channel RAID SCSI 15K for the Oracle database.

Collapse
3: Re: Performance impact of new dotlrn mount for every instance of a class (response to 2)
Posted by Peter Alberer on 08/19/03 11:53 PM
thanks for the info malte, i will give the default setup a try.