Hyper V Performance Question - compare to Bare Metal

plq
plq used Ask the Experts™
on
I've tried googling for this and not found an answer

I have 8 HP Servers with Quad core CPUs, 16 GB RAM, and RAID 5 SCSI disks 6x73GB giving me 3x73GB disk space all on the C: drive

We standardised on baremetal in the configuration as there was no need to split the servers.

The apps run a custom https web service with a sql server backend. About 75% of cpu and disk activity is taken up by sql. About 15% by our application, and the remaining 10% by IIS and the OS. Sql Server uses 14-15GB RAM and the rest is OS and our app (small)

Now I'm wondering if I created a single HyperV instance within each server, and moved the software into there, how much performance degradation would I be likely to experience ? - so replacing one bare metal server with one hyperv vm hosted on that same one bare metal server. Hope that makes sense..

I kind of feel that Sql would perform much worse in a virtualised environment but thats more gut feeling than knowledge

The reason - I want to make the VM's moveable for DR and failover reasons. ATM we maintain identical servers and just move the sql databases around for DR, but I feel there would be added resilience if we could move the whole machine.

thanks
Comment
Watch Question

Do more with

Expert Office
EXPERT OFFICE® is a registered trademark of EXPERTS EXCHANGE®

Commented:
You can do what your wanting, in fact, whatever you want if you will allow reads/writes from an ioDrive (or similar tech). For the money you've spent so far I think you could afford it.
http://www.fusionio.com/products/iodrive/
plq

Author

Commented:
Thanks for this but I can only change the software and add a hyperv guest. I cannot buy hardware.

I just wanted to know how much worse the performance would be in a virtual setup..

thanks
Commented:
Hello plq,

Virtualization in General is not a key solution to host services that consume very high resources such as MS SQL Server, however, there are some cases where you might consider virtualizing your environment depending on your need.

The performance degradation is not something that can be measured on the fly, for instance, HyperV supports a maximum of 4 vCPUs per VM, which means that you will lose processing power on the VM if the source host that was running SQL had more than a QuadCore CPU, disk performance might be affected if you use VHDs instead of Physical Drive for your VM and so on. we can provide performance optimization tips, but nothing is guaranteed in the real environment.

There is no exact metric that can be used to evaluate how much performance you will lose unless you put it into the test.
you might want to do a pilot and measure the performance, i guess its worth the try.

Hope this helps,

plq

Author

Commented:
thank you.

Do more with

Expert Office
Submit tech questions to Ask the Experts™ at any time to receive solutions, advice, and new ideas from leading industry professionals.

Start 7-Day Free Trial