Re: Evaluation
Adam Tauno Williams <[email protected]> Thu, 14 May 2009 12:50:05 -0400
| Newsgroups | gmane.comp.cms.opengroupware.user |
|---|---|
| Message-ID | <1242319805.5291.2.camel@linux-m3mt> |
On Wed, 2009-05-13 at 21:46 -0400, Dirk H Bartley wrote: > > > My testing with zideone beta yielded 2 results which make proposing > > > using zideone impractical as it stands. > > > 1. Connecting to enterprises where I had more than 3K enterprises, the > > > server's zidestore process went to 99 percent processor usage just from > > > connecting to zideone, steady until outlook was stopped. Even after > > > zideone had all of the enterprises in view. That was with just one > > > outlook connection and I want about a hundred. > > > > While that seems a bit high it isn't abnormal for the *VERY* first sync. > > [although I thought B3 had a pacing algorythm to avoid pummeling > > servers] The ZideStore server has to basically render the entire DB > > contents as vCard (VCF) for the GroupDAV clients (TB/ZideOne/Evolution, > > etc...) BUT this only happens the very first time; after that the > > content is served from the cache and throughput should be pretty good. > > If you look in /var/lib/opengroupware.org/documents you'll see all the > > VCF files once you've sync'd. You can increase this efficiency even > > more by hashing the cache via the LSVCFCachePath/LSHashVCFCache defaults > > (should be in the ZideStore chapter of WMOGAG). > Searching in wmogag, the string LSVCF does not exist, at least in the > version I have. There is SxCachePath on Pg58. > /var/lib/opengroupware.org/documents does seem to be the path, filled > with files. LSHashVCFCache default does certainly change the look of > the directory, I've pushed a new version, look like I added those since my last "publication". <http://docs.opengroupware.org/Members/whitemice/wmogag/file_view> > but the processor will still stay pegged at 99 until I > stop outlook, I mean for hours and hours, non-stop cpu peggage. That seems like a long time; has Outlook completed sync-ing? I assume the CPU pegging is done by ogo-zidestore and not PostgreSQL? I do see high-cpu utilization when I first start Outlook but it slacks off once everything is in sync, and fairly quickly if the cache is full. But even then the server remains responsive; the ngobjweb module stacks up the requests and shoves them though ZideStore as fast as it can. -- OpenGroupware.org Users [email protected] http://mail.opengroupware.org/mailman/listinfo/users