- From: Anselm Baird-Smith <abaird@w3.org>
- Date: Wed, 26 Jun 1996 19:34:37 +0500
- To: alex@access.com.au (Alexandre Rafalovitch)
- Cc: www-jigsaw@w3.org
Alexandre Rafalovitch writes: > At 6:13 PM on 26/6/96, Anselm Baird-Smith wrote: > > > Doug Marker writes: > > > Re the peroformance stats - we notice they are high to start with but > > > after a day or so drop markedly - we assume this is due to the caching > > > techniques used ? > > > > Hum...I would be glad to get more infos about this. I know that with > > the Sun's JDK they are troubles with the garbage collector, and it is > > safe to run jigsaw with the asyn GC turned of (java -noasyncgc > > ...). However, I uderstand that you are not using Sun's JDK ? Could > > you tell us about what the Statics resource says about free memory ? > > Could it be because java never unloads class definitions even if it has no > instances of that class in memory? Then, as people access more and more > different resources/classes they are loaded into memory and never unloaded. > Until Jigsaw is shutted down of course. If that is the reason for > performance drop, then it is not a fault of Jigsaw, but of Java runtime. The problem ha to do with the way the async GC interacts with the sync GC. For more infos have a look at: http://www.w3.org/pub/WWW/Jigsaw/User/Introduction/performance.html [or something in this directory] > About garbage collecter, maybe Jigsaw should call it syncronously in worst > cases with System.gc()? > Ignore this comment , if it already does it. It doesn't, although it should probably at least call the finalizer at some points... Anselm.
Received on Wednesday, 26 June 1996 19:34:41 UTC