[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-API] XCP 1.5 eta

On 06.01.2012 18:54, Konrad Rzeszutek Wilk wrote:
We running cloud on XCP, I can say, that cpu performance is
negligible factor for current hypervisors state. The main problem
starts around network and disk performance, and balance of
features/speed of them.

Our virtual machines boots within this time (10s and less) and I
can't say this is somehow critical question. More important is, f.e.
footprint on dom0 for network performance. I somehow unhappy of XCP
performance - I getting only 5-7GiB/s with about 100% load of one
dom0 vCPUs. I've prefer to see something around 8-9GiB/s with much
Does it change if you make dom0 have more vCPUs?

At certain level, yes. But linear performance (per VM) will been kept in performance of single CPU...

lower load, but not sure if any other system with bridge level
aggregation and live migration could perform better.

Same for disk IO - XCP do have relatively slow ISCSI, performing
only around 300-500MB/s (instead expected over 1GB/s), but, again, I
saw nothing better right now.
1GB being what you saw on baremetal? Or is that what your line speed is?

Baremetall shows something like 1.2-1.5GiB/s (bunch of disks in huge raid).
One problem with 'slow XCP' may be within HVM - which is lame mode
for lame windows-based appliances.
Are you using the PV drivers in your HVM guest?
I don't use windows and don't use HVM at all. :)

xen-api mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.