Virtual Performance

RE: Using Perfmon For Accurate, ESX Performance Counters

praveen — Mon, 09 Nov 2009 22:22:36 GMT

MartinWi,

The counters we expose through Perfmon are built on top of the guest SDK. We currently don't have an equivalent tool on Linux that mirrors the ease of use of Windows Perfmon. However, you can probably leverage the guest SDK and write a tool/wrapper to get to the "accurate" CPU counters on Linux guests (w/ tools installed). See the documentation at http://www.vmware.com/support/developer/guest-sdk/. Both the 3.5 and 4.0 versions have the same counters that you are interested in.

RE: Using Perfmon For Accurate, ESX Performance Counters

MartinWi — Thu, 29 Oct 2009 17:25:25 GMT

This is great stuff. Makes it easy with snmptools to get the values via snmp, and graph usage in Mhz via for example cacti.
I wonder though if there is anything similar to use with Linux(i.e. get the usage in Mhz from the virtual machine somehow)? The documentation about VMware Tools doesn't exactly provide much info.

RE: Using Perfmon For Accurate, ESX Performance Counters

Itzikr — Wed, 23 Sep 2009 09:40:39 GMT

hi,
are you aware of any bug, when even after upgrading to vsphere and after upgrading the vmware tools and the virtual hardware to ver 7, the counters are still missign

Itzik Reich

Solutions Architect

VCP,VTSP,MCTS,MCITP,MCSE,CCA,CCNA

EMC²

where Information Lives

If you find this information useful, please award points for "correct" or "helpful".

RE: Using Perfmon For Accurate, ESX Performance Counters

drummonds — Wed, 09 Sep 2009 17:24:43 GMT

I have again uploaded the DLL, for those that are interested.

Scott

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

RE: Four Things You Should Know About ESX 4's Scheduler

Ken.Cline — Sat, 22 Aug 2009 18:55:54 GMT

Hey Scott,

Great article, as usual. One comment on Item 4...if you don't understand how contention works into the equation, you could mistakenly assume that it is better to create a dual vCPU VM than a single vCPU VM - you have more "options" for placement.

I can hear it already:

Me: "You should always create your VMs with a single vCPU and scale up if you experience performance problems"

Customer: "But Scott Drummonds' article says a 2 vCPU VM has a 50% better chance of being scheduled than a single vCPU VM..."

Me: "OK, let's have a discussion about contention and scheduling opportunities. Where's the whiteboard?"

Thanks!
KLC
Ken Cline
VMware vExpert 2009
VMware Communities User Moderator
Blogging at: http://KensVirtualReality.wordpress.com/

RE: Four Things You Should Know About ESX 4's Scheduler

jasonboche — Sat, 22 Aug 2009 04:13:10 GMT

Nice post. One thing on item 4 (cells). People are going to be on ESX 3.x for a while and 6 core processors are already here. I think it would be a nice addition if you would add to your diagrams VM placement on 6 core sockets with an additional note about changing the default cell size to accommodate 6 core procs and the impacts of doing so and not doing so. Also add a note on how Intel's modern HT fits into the picture.

Thanks,
Jas

Jason Boche, vExpert
boche.net - VMware Virtualization Evangelist
VMware Communities User Moderator
Minneapolis Area VMware User Group Leader

RE: Using Perfmon For Accurate, ESX Performance Counters

LMSSML — Mon, 17 Aug 2009 21:01:48 GMT

thanks drummonds!

RE: Using Perfmon For Accurate, ESX Performance Counters

pwcvmware — Mon, 17 Aug 2009 17:00:07 GMT

thanks Scott!

RE: Using Perfmon For Accurate, ESX Performance Counters

drummonds — Mon, 17 Aug 2009 16:11:54 GMT

I have re-uploaded the file. Enjoy!

Scott

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

RE: Using Perfmon For Accurate, ESX Performance Counters

LMSSML — Mon, 17 Aug 2009 14:44:48 GMT

Link is broken

Probably cleaned.

Is it possible to re-upload ?

Thanks in advanced.

RE: Using Perfmon For Accurate, ESX Performance Counters

pwcvmware — Mon, 17 Aug 2009 10:02:21 GMT

cannot download the tool. Maybe the file was cleaned up again?
Do you have another location where I can download the file?

Ronny

RE: Using Perfmon For Accurate, ESX Performance Counters

drummonds — Mon, 03 Aug 2009 17:32:53 GMT

The temporary location that we are hosting the file from gets cleaned regularly. I have re-uploaded the file.

Scott

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

RE: Using Perfmon For Accurate, ESX Performance Counters

dmw1234 — Fri, 31 Jul 2009 13:57:28 GMT

link to the tool doesn't work. is the tool still available?

RE: SQL Server Performance Problems Not Due to VMware

drummonds — Thu, 09 Apr 2009 22:55:53 GMT

Thanks for the additional information, Brian. I'd agree with both of your comments, with caveats. As you can see from above, the SQL priority boost did deliver additional gains to performance. Our experience with real deployments is that the gain we measured may not translate into the same benefits in production environments. This being due to the fact that increased priority of the SQL process can cause problems with other processes that could be integral to your architecture's efficiency.

However, I will note that the purpose of this blog was to point out that priority boost could generate up to 5% gain, but this is trivial in the context of performance problems that are halving SQL performance or worse.

RE: SQL Server Performance Problems Not Due to VMware

KBrianKelley — Thu, 02 Apr 2009 15:46:41 GMT

I'm a big fan of virtualization and we've sought to virtualize where it made sense (and it usually does). However, some of the things you're saying with regards to SQL Server configuration I disagree with. I would in no way say that SQL Server priority boost is a Microsoft best practice for SQL Server configuration. In fact, the truth is it's not. Reference:

http://support.microsoft.com/kb/319942

"Based on actual support experience, you do not need to use priority boost for good performance. If you do use priority boost, it can interfere with smooth server functioning under some conditions and you should not use it except under very unusual circumstances. For example, Microsoft Product Support Services might use priority boost when they investigate a performance issue."

And large pages only come into play when you have x64 or ia64 installations with at least 8 GB of RAM (reference: http://blogs.msdn.com/slavao/archive/2005/02/11/371063.aspx). So for smaller installations, that's not available and cannot be suggested as a best practice.

RE: DPM Power/Performance Video

drummonds — Thu, 05 Mar 2009 18:53:37 GMT

DPM will be a fully supported feature in the next version of VI that is coming out this year.

RE: DPM Power/Performance Video

ykinfo — Thu, 05 Mar 2009 08:18:32 GMT

when the DPM feature will be release?