Hello all, I have been running different benchmarks on paravirtualized domU. I have run into strange performance drops with CPU intesive benchmarks (64-bit BurnInSSE & 64-bit Linpack test). With Linpack test the performance of Xen is 90% of hardware where KVM is only about 1-2% worse than hardware. This is something new to me as I am used to better performance with older Xen versions. I have a Dell R410 with two Xeon E5520 processors and I am running Ubuntu Lucid on both domU and dom0. The Xen settings are pretty much default, though I did try varying credit scheduler weights without any success. Has anybody else noticed the same or have I just forgot something essential from my kernel? Best regards, Jukka Kommeri _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Sat, Oct 15, 2011 at 4:21 AM, Kommeri Jukka <jukka.kommeri@aalto.fi> wrote:> Hello all, > > I have been running different benchmarks on paravirtualized domU. I have run > into strange performance drops with CPU intesive benchmarks (64-bit > BurnInSSE & 64-bit Linpack test). > > With Linpack test the performance of Xen is 90% of hardware where KVM is > only about 1-2% worse than hardware. This is something new to me as I am > used to better performance with older Xen versions. >Linux 3.0 is not performance optimized yet. There are some known bugs. Many of which are fixed in newer versions. See: http://wiki.xen.org/xenwiki/Linux_30_bugs For more hints on tweaking things see: http://wiki.xen.org/xenwiki/XenParavirtOps Can you provide us with more details of your setup so that we can give you more suggestions? It would be better to be testing the newest Linux kernel, since a lot of changes have gone in since Linux 3.0> I have a Dell R410 with two Xeon E5520 processors and I am running Ubuntu > Lucid on both domU and dom0. The Xen settings are pretty much default, > though I did try varying credit scheduler weights without any success. > > Has anybody else noticed the same or have I just forgot something essential > from my kernel? > > Best regards, > Jukka Kommeri > > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xensource.com > http://lists.xensource.com/xen-users >-- Todd Deshane http://www.linkedin.com/in/deshantm http://www.xen.org/products/cloudxen.html http://runningxen.com/ _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Konrad Rzeszutek Wilk
2011-Oct-24 19:53 UTC
Re: [Xen-users] Xen 4.1 and 3.0 kernel CPU performance
On Mon, Oct 17, 2011 at 11:33:27AM -0400, Todd Deshane wrote:> On Sat, Oct 15, 2011 at 4:21 AM, Kommeri Jukka <jukka.kommeri@aalto.fi> wrote: > > Hello all, > > > > I have been running different benchmarks on paravirtualized domU. I have run > > into strange performance drops with CPU intesive benchmarks (64-bit > > BurnInSSE & 64-bit Linpack test).What happens if you run PVonHVM domU? Meaning you boot the guest as if it is HVM? This is 3.0 DomU and 3.0 Dom0?> > > > With Linpack test the performance of Xen is 90% of hardware where KVM is > > only about 1-2% worse than hardware. This is something new to me as I am > > used to better performance with older Xen versions.Hmm..> > > > Linux 3.0 is not performance optimized yet. There are some known bugs. > Many of which are fixed in newer versions. > > See: http://wiki.xen.org/xenwiki/Linux_30_bugs > > For more hints on tweaking things see: > http://wiki.xen.org/xenwiki/XenParavirtOps > > Can you provide us with more details of your setup so that we can give > you more suggestions? > > It would be better to be testing the newest Linux kernel, since a lot > of changes have gone in since Linux 3.0<nods> Did you upgrade Jukka? _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users