[ngw] Antw: Re: GroupWise 2012 on Windows 2008R2 Slow

Jeffrey Sessler jeff at ScrippsCollege.edu
Wed Aug 14 18:36:33 UTC 2013


It some cases, this is going to be true e.g. SQL server, but it will depend on the VM host. At the point the host is busy waiting for CPUs to free up, the performance on the 2vCPUs can become worse than a 1vCPU.
 
If you do see a performance increase, I'd be curious to know if the guests are using the vmxnet3 network and paravirtual SCSI (pvscsi) drivers. If not, the extra vCPU may be making up for the lower-performing network/disk driver options.
 
Jeff

>>> On Wednesday, August 14, 2013 at 10:22 AM, in message <520B84B00200001600006B60 at k12group.net>, "Keith Larson" <klarson at k12group.net> wrote:

I have not found this to always be true.  I build all vm's with a single cpu and single core, but I have had some OES2SP3 servers that really started performing better when this was increased to 2 cpu's.


Keith Larson
Franklin Computer Services - K12group
klarson at k12group.net
(614) 561-4887 (mobile)



>>> "Thomas Stimper" <tstimper at nmedv.de> 8/14/2013 12:45 PM >>>
> The #1 issue I find in virtual deployments is the over use of virtual CPUs. 
> Guests deployed with 2-4 virtual CPUs, when they really should only have one. 

Agree. We learned it this week that  VMs with 1vCPU are running faster that VMs with 2 or more vCPUs.

Thomas




>>> "Jeffrey Sessler" <jeff at ScrippsCollege.edu> schrieb am Mittwoch, 14. August
2013 um 18:26:
> I'd start by looking at the performance graphs for the VM host and GW guest. 
> This will provide a good amount of information on what's happening with 
> memory, cpu, and i/o. It possible that you'll see pressure on the host yet 
> the GW guest is AOK. If that's the case, you need to see what's impacting it. 
> Go back and look at the day/time when this slowdown occurred and see if 
> anything jumps out.
>  
> GW doesn't do a lot of its own caching, so it counts on the OS file caching. 
> In a VM environment, the VMWare tools may reclaim file cache memory if the 
> host is under pressure and another guest needs it. This could result in the 
> GW host slowing down since the agents are having to go back to disk.
>  
> CPU and Memory reservations are one way to protect the resources allocated 
> to the GW guest, but it could be at the expense of other guests.
>  
> The #1 issue I find in virtual deployments is the over use of virtual CPUs. 
> Guests deployed with 2-4 virtual CPUs, when they really should only have one. 
> This puts a huge load on the host since it has to free up concurrent CPUs for 
> those guests.
>  
> Jeff
> 
> 
>>>> On Tuesday, August 13, 2013 at 12:29 PM, in message 
> <520A50D8020000EC00058141 at groupwise.williamsburgva.gov>, "Mark BARHAM" 
> <mbarham at williamsburgva.gov> wrote:
> 
> We currently run a single Groupwise 2012 Post Office/Primary domain on a 
> single Windows Server 2008R2 (VMWare).  Windows is fully patched and 
> GroupWise is patched to SP1.  Secondary domain/GWIA/Webaccess run on a 
> separate Windows 2008R2 server.  Post office is approximately 275 users and 
> is about 250GB in size.  Users are a mix of GW* and 2012 and all connect via 
> Client/Server.
> 
> This setup has been in place for several years and has worked well until a 
> few months ago.  Now, everything is just very slow at random times of the day 
> and routinely ends up in a "Not Responding" mode on the clients.  Kill it, 
> log back in to GroupWise and all is well.  There is no client side firewalls 
> or any internal filtering between client and server and DEP is turned off at 
> the server.  Can anyone provide any direction and/or guidance as to where I 
> can look to next to resolve this?  Thanks
> 
> 
> 
> 
> 
> 
> 
> City of Williamsburg
> www.williamsburgva.gov 
> 
> Facebook
> 
> Twitter
> 
> YouTube
> 
> City411

_______________________________________________
ngw mailing list
ngw at ngwlist.com
http://ngwlist.com/mailman/listinfo/ngw


More information about the ngw mailing list