Phil Leinhauser
Tue, 09 Feb 2010 09:39:33 -0800
Sorry 'bout that. Never used threaded view...
So, I restarted this one and added below... Sorry, Phil. I posted reply to Jake before I read this. BTW, when you change a thread, please create a new message (copy/paste content as appropriate) instead of replying and changing the subject. When you look at messages in threaded view (nice feature btw), this isn't a new thread, only a changed subject. :( Common mistake. Phil Leinhauser wrote: > So let's change the thread and explore yours a bit... > > I'd be interested in where virtual QMTs choke. Is it IO, proc, mem? > Maybe too many other guests? > > While I only have about 1/2 of your 500 user choke point, I get very > good performance from my VM server based QMT. I did have issues before > I found some of the tuning points that I'm documenting in the wiki. > I've been running some sort of VM in my small business since the days of > GSX and in work I have a very large install of ESX / vSphere. I do know > things like heavily used databases and the like don't run too well > virtualized. I'd be interested to see if the QMT limit is something > that can be overcome? Me too. I think/feel/hope there's a way to improve on that. FWIW, I talked with a company a year or so ago that was running Oracle in virtual guests, and they had managed to obtain satisfactory performance. They didn't go into specifics, so I don't know how they did it. Can be done though I believe. > I saw your discussion a few days ago with Eric about slow virtual > performance but you said that was with something other than VM? > That was one Xen server. I don't know about the others he was working on > though. We did have a couple Oracle databases on our ESX 3.5 but they seemed to choke. I was in the middle of building the infrastructure so I couldn't help him (Not that I would have been much help with Oracle). Is there anything like a load tester for QMT to mimic the 500 user limit?