RE: [Xen-devel] VMware multi-nic performance tech note

["Ian Pratt" <Ian.Pratt@xxxxxxxxxxxx>]

> VMware's published a paper on ESX vs. Xen (Enterprise) multi-nic
> performance:
> 
> http://www.vmware.com/pdf/Multi-NIC_Performance.pdf
> 
> Basically, it tests network performance from a Windows 2003 
> guest under Xen and VMware, and compares the results.
> Just curious if there's any work in progress that would 
> improve this? 

There certainly is -- it was discussed at the summit:
 
 
http://www.xensource.com/files/xensummit_4/xen-ny-summit-smartnic_Pratt.
pdf

Their previous benchmarking paper was a good spur to action resulting in
us implementing optimizations that mean we now meet and in many cases
beat the ESX results for the benchmarks they selected. 

Soon they'll figure out that goading us on the handful of benchmarks
they might actually win on is a really bad idea as we're able to respond
more quickly than they can possibly imagine :) 
 
> I'd also love to see the test repeated with a 
> Linux PV guest versus the fastest way to run a Linux guest on 
> VMware to see how the performance is there.  :)

Linux certainly wouldn't be VMware's benchmark battleground of choice :)

> I'd aso be very curious to hear if VMware consulted with 
> XenSource to ensure that their benchmarking configuration was 
> optimal, as VMware requires you to do to publish benchmarks on ESX:
> 
> http://www.vmware.com/pdf/benchmarking_approval_process.pdf

AFAIK there was no consultation, but I have no reason to doubt the
results: we haven't put a lot of effort into three NIC systems -- Yet. 

Ian

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel