Currently, we run about 10 VM's on a 2003 x64 box.. with 12GB of memory..
this box is a dual CPU quad core xeon 2.4GHZ box with a SATA2 32mb cache
raid6 array, consisting of 8, 500GB drives.. Our network has about 40 users
and 35 machines roughly.. i dont think our sql box gets hit that frequently
either..



The performance Passmark rating of this box on the D drive array test is
about 3850 (This is with passmark software located at
http://www.passmark.com/products/pt.htm there is also a free trial)

For the Virtual Servers, even with 10 of them.. i get about 565 for an
average on each one (this is the equivalent of a dual core system)..
technically (i think) this is about 90% slower than the physical machine, if
you go by passmark ratings.. which maybe i shoudlnt ..

Things run fine.. i run virtually: Office Comm server, both Domain
controllers (anyone else doing DC's completely virtually?) with one on this
server and another DC on another virtual server that only runs a DC, i also
run Virtual Machine manager (guessing i wont need this with hyperV).. I run
MS Dynamics Great Plains, a sql 2005 box with 7 virtual instances..
sharepoint virtually.. Team Foundation server.. Configuration Manager..
antivirus trendmicro box.. etc..

In the case of the virtual SQL, i've installed the sql portion of say the
configuration manager or dynamics in these instances (seperately).. anyone
else doing this?

So my question is..

How is the performance of Hyper-V.. either barebones hyper-V or the software
overlayed Hypervisor..

Would anyone be willing to run the passmark test on the Host machine then on
a virtual server in either of these cases?

Is it really 85% that of the host machine with the barebones.. what is it
with the software version ontop of 2008 server?

How is everyone else's performance with VS 2005? Anyone running actual file
systems/non-passthrough inside a VM rather than on a host?



Thanks in advance..

In preparation for our user group event on September 29, we may want to show
some performance data for Hyper-V. I could run some of these benchmarks on
the systems we will be demonstrating. However, I won't be able to run these
until sometime in September, due to other projects in the queue ahead of this.

My servers are dual-Xeon, quad-core with 32GB of memory, with internal SAS
drive array and connections to iSCSI and Fibre Channel disk storage. What
applications are you running on your VMs? How much memory are you allocating
to each of your VMs?

"markm75" wrote:

> Currently, we run about 10 VM's on a 2003 x64 box.. with 12GB of memory..
> this box is a dual CPU quad core xeon 2.4GHZ box with a SATA2 32mb cache
> raid6 array, consisting of 8, 500GB drives.. Our network has about 40 users
> and 35 machines roughly.. i dont think our sql box gets hit that frequently
> either..
>
> The performance Passmark rating of this box on the D drive array test is
> about 3850 (This is with passmark software located at
> http://www.passmark.com/products/pt.htm there is also a free trial)
>
> For the Virtual Servers, even with 10 of them.. i get about 565 for an
> average on each one (this is the equivalent of a dual core system)..
> technically (i think) this is about 90% slower than the physical machine, if
> you go by passmark ratings.. which maybe i shoudlnt ..
>
> Things run fine.. i run virtually: Office Comm server, both Domain
> controllers (anyone else doing DC's completely virtually?) with one on this
> server and another DC on another virtual server that only runs a DC, i also
> run Virtual Machine manager (guessing i wont need this with hyperV).. I run
> MS Dynamics Great Plains, a sql 2005 box with 7 virtual instances..
> sharepoint virtually.. Team Foundation server.. Configuration Manager..
> antivirus trendmicro box.. etc..
>
> In the case of the virtual SQL, i've installed the sql portion of say the
> configuration manager or dynamics in these instances (seperately).. anyone
> else doing this?
>
> So my question is..
>
> How is the performance of Hyper-V.. either barebones hyper-V or the software
> overlayed Hypervisor..
>
> Would anyone be willing to run the passmark test on the Host machine then on
> a virtual server in either of these cases?
>
> Is it really 85% that of the host machine with the barebones.. what is it
> with the software version ontop of 2008 server?
>
> How is everyone else's performance with VS 2005? Anyone running actual file
> systems/non-passthrough inside a VM rather than on a host?
>
>
>
> Thanks in advance..
>
>

"Dennis29" wrote:

> In preparation for our user group event on September 29, we may want to show
> some performance data for Hyper-V. I could run some of these benchmarks on
> the systems we will be demonstrating. However, I won't be able to run these
> until sometime in September, due to other projects in the queue ahead of this.
>
> My servers are dual-Xeon, quad-core with 32GB of memory, with internal SAS
> drive array and connections to iSCSI and Fibre Channel disk storage. What
> applications are you running on your VMs? How much memory are you allocating
> to each of your VMs?
>

That would be great, particularly comparison of the same system on Virtual
Server vs Hypervisor.. but either way that is good info..

I allocate anywhere from 1GB to 4GB depending on the application.. I isolate
applications per individual VM's..

IE: Configuration Manager 2007 on a VM called VMCM01 (say 1 or 2GB
memory).. then all sql instances on a sql VM called VMSQL01 with 4GB memory..

Then another with antivirus software.. call it VSAV01 with say 1GB of memory..

usually i do fixed discs, some i do dynamic.. though with our 40 user load,
i could see a difference in performance, even on the sql box.

Curious.. what type of motherboard are you using giving you 32GB? Our next
board was going to be one that could do more than 16GB.. the Supermicro
motherboard we use now cant do more than 16GB.

Thanks

These servers are Dell PowerEdge 2900 that support up to 48GB RAM. I
currently have 32GB but may increase them to 48GB.

Are you running any 64-bit VMs or applications? I usually run 64-bit
all-around and can allocate more than 4GB to a VM if needed. SQL 2005 and
Exchange 2007 like lots of memory.

"markm75" wrote:

>
>
> "Dennis29" wrote:
>

> > In preparation for our user group event on September 29, we may want to show
> > some performance data for Hyper-V. I could run some of these benchmarks on
> > the systems we will be demonstrating. However, I won't be able to run these
> > until sometime in September, due to other projects in the queue ahead of this.
> >
> > My servers are dual-Xeon, quad-core with 32GB of memory, with internal SAS
> > drive array and connections to iSCSI and Fibre Channel disk storage. What
> > applications are you running on your VMs? How much memory are you allocating
> > to each of your VMs?
> >

>
> That would be great, particularly comparison of the same system on Virtual
> Server vs Hypervisor.. but either way that is good info..
>
> I allocate anywhere from 1GB to 4GB depending on the application.. I isolate
> applications per individual VM's..
>
> IE: Configuration Manager 2007 on a VM called VMCM01 (say 1 or 2GB
> memory).. then all sql instances on a sql VM called VMSQL01 with 4GB memory..
>
> Then another with antivirus software.. call it VSAV01 with say 1GB of memory..
>
> usually i do fixed discs, some i do dynamic.. though with our 40 user load,
> i could see a difference in performance, even on the sql box.
>
> Curious.. what type of motherboard are you using giving you 32GB? Our next
> board was going to be one that could do more than 16GB.. the Supermicro
> motherboard we use now cant do more than 16GB.
>
> Thanks