Phoronix IRC Log: 2009-04-14
schestowitz: "The Phoronix Test Suite is our GPLv3..." well done, Michael
schestowitz: happy (belated) Easter
vadi2: anyone here a vm expert? these iozone results from inside the guest are baffling: http://www.ubuntu-pics.de/bild/12476/screenshot_100_710__SjjmGH.png and http://www.ubuntu-pics.de/bild/12477/screenshot_100_711__l9R3BN.png
Kano: vadi2: i dont think that benchmarks inside a vm are correct
Kano: vadi2: because they are always cached - there is no way to bypass hd cache at all
vadi2: no, the issue is that this one particular vm host has great performance and the guest has miserable performance.
vadi2: not sure, still trying to dig to the bottom of this.
Kano: do you use vbox?
Kano: with addons it is faster to use virtual folders instead of running an external samba server
Kano: but it is impossible that you get the native speed in a guest
vadi2: no, the data is all local in the vm
vadi2: it's a db inside the vm
Kano: well in a growing image it is slower than in a fixed size one
vadi2: yes, that could be a reason. I'll try it with fixed
redeeman: Kano: vm software should NOT cache
redeeman: but vadi2 that makes perfect sense
redeeman: the CoW images of vbox are much slower than real
redeeman: use a raw block device either as a preallocated file, or better yet, a raw disk
redeeman: and you'll get far better performance
redeeman: but also, vbox sucks :)
vadi2: a raw disk?
redeeman: yes, aka, not a CoW file
vadi2: and no, linux sucks, the whole reason I have to use a vm is because of the horrible io bug. grr.
vadi2: yeah so a fixed-size one. ok
redeeman: well.. i don't really think you should be talking about what sucks, when you don't know more about it :)
vadi2: I know about its effects :)
GNU\colossus: what io bug?
Kano: michaellarabel: the md5sum of warsow seems to be bad in pts