March 29th, 2007

Benchmarking saga continues

Benchmarking of virtualization technologies has apparently become a popular news subject.
XenSource has just published a paper that compares VMware ESX 3 and XenSource  Enterprise 3.2 – as already reported by and Individual results have been already available, but VMware kindly allowed XenSource to publish the results in a single paper, side-by-side. The result is that in terms of performance, XenSource is as good or slightly better than ESX. Also, the more virtual CPUs are added, the more confident XenSource leads.
As you might remember, VMware originally compared commercial Windows-optimized version of their product with version of Xen, which was not intended for commercial use and not optimized for Windows.
Now, couple of days ago, InfoWorld compared four desktop virtualization products and found that in most tests Parallels 2.2 is faster than beta of VMware Workstation 6.0. Here VMware says that it’s not correct to put a commercial optimized version of Parallels Desktop against VMware Workstation Beta 6.0, which is of course not intended for commercial use and not optimized.
Quod licet Iovi, non licet bovi
If you are wondering how this is all related to Virtuozzo, look at what Kir wrote in his OpenVZ blog You can see the results for yourself here (German original) and here ( fragment translated to English). As you might now, OpenVZ is the foundation of Virtuozzo for Linux distributed as open source product. To summarize:
·         OpenVZ is on par with Xen for CPU-intensive tasks
·         OpenVZ is a bit better than Xen in networking
·         OpenVZ is somewhat or much better than Xen in I/O and IPC
And finally – thanks for the newshere is the comparison of the several popular server virtualization technologies – VMware, Xen, UML and Vserver. Look for yourself – it’s a pretty interesting read. I will try to contact these guys and ask them to add Virtuozzo in the mix, and then I’ll get back with more detailed analysis.