We were running into some performance issues where people seemed to be waiting for responses now and then.
It turned out there had been a cock up with the WLS ExecuteThreads setting in config.xml. For some reason they weren’t set, so we had the default value of “15″. This wasn’t enough, so we had to bump up the amount.
This is always a bit of an art. You can’t just put them up to a huge amount as then you have a bunch of threads that are taking up memory, and context switching starts to bring your application to a crawl. Via a smart guess, and a little tweaking with trial and error, and we had the right values again. For now.
Luckily, WLS 9 (code named Diablo), has an interesting new feature. It can be a lot smarter with the various settings such as thread pools, accept backlogs, etc etc. It can be a lot smarter than us, and it can change things based on the CONTEXT of the running server.
If this really works I will be a happy camper. In consulting gigs I have found that when people working on the project make changes to these values they usually cause MORE problems!
Here is to the core WLS team. Good luck in getting it right for Diablo!
