Performance & VMmark : VMkernel Scheduler : Comments

RE: VMkernel Scheduler

Ken.Cline — Thu, 30 Oct 2008 04:36:12 GMT

Scott,

Thanks for this document. I think it will be very helpful over time. Keep up the good work!

KLC
Ken Cline
Technical Director, Virtualization
Wells Landers
TVAR Solutions, A Wells Landers Group Company
VMware Communities User Moderator

RE: VMkernel Scheduler

alex0 — Tue, 23 Jun 2009 10:42:20 GMT

Hi Scott,

Can you please update this docnote to explain how the cell system works in ESX4?

Also this OFFICIAL VMWARE document, http://www.vmware.com/pdf/tips_tricks_infrastructure_services.pdf, on Page 4, states that there are 1820 ways of scheduling a 4 vCPU workload on a 16 core ESX3 system, can you please confirm that this is patently incorrect.

Regards,
Alex

RE: VMkernel Scheduler

drummonds — Thu, 25 Jun 2009 10:39:10 GMT

The count of 1820 possibilities considers individual vCPU mappings to specific cores. This means that a system with only two cores has two ways to place a 2-way VM. But there is no performance difference in these two positions so I eliminate the second choice, which flips the vCPU to CPU mapping. I think that both documents are correct.

I'll update this document for ESX 4 once we release our vSphere scheduler document. In short, there is no longer a cell with VMware vSphere.

Scott

More information on my communities blog and on Twitter:
http://communities.vmware.com/blogs/drummonds
http://twitter.com/drummonds1974

RE: VMkernel Scheduler

thinks2much2 — Tue, 07 Jul 2009 09:01:48 GMT

Hi Scott,

Do you have an ETA on the vSphere scheduler document? Something definitive that shows the benefits of vSphere would be very much appreciated.

Many thanks,

Kevin

RE: VMkernel Scheduler

drummonds — Tue, 07 Jul 2009 16:33:10 GMT

Two weeks.

More information on my communities blog and on Twitter:
http://communities.vmware.com/blogs/drummonds
http://twitter.com/drummonds1974

RE: VMkernel Scheduler

aurimask — Fri, 31 Jul 2009 11:04:59 GMT

Where to get information of how VMKernel works if VM are assigned of more resources than one physical ESX has? For example cluster is 2 ESX each 2ghz quad core and you reserve 10 ghz for VM in a cluster.

RE: VMkernel Scheduler

drummonds — Fri, 31 Jul 2009 16:08:59 GMT

A virtual machine can only use 100% of the number of vCPUs provided to it. Increasing limits above this number--if this is even possible--would not provide a performance gain.

Scott

More information on my communities blog and on Twitter:
http://communities.vmware.com/blogs/drummonds
http://twitter.com/drummonds

RE: VMkernel Scheduler

aurimask — Fri, 31 Jul 2009 19:06:44 GMT

So it is not possible for virtual machine have more cpu resources than physical ESX server on which VM is residing has?