We are in the process of evaluating how VMWare virtualization may fit into our IT infrastructure. With the forthcoming purchase of a SAN for PACS storage, one major component of the virtualization framework is accounted for.

The aspects that are appealing with VMWare are:

– disaster recovery. Having the ability to start up a virtual machine from a second host server is great. The possibilities for seamless transitions with VMotion are fascinating.

– resource management. We currently have numerous servers that sit idle through much of the day. Being able to better manage our server resources and consolidate onto fewer server boxes will ease the management overhead and costs of our server room.

There are other benefits to be sure, but those two are foremost at this point.

If we proceed with VMWare, it would be with the ESX product, as it has many features that the free Server product lacks. However, the Server software is allowing us to test the waters, and we are liking what we see so far.

Being able to quickly deploy test environments has been eagerly adopted by members of our IT team. We are currently testing and configuring our new scheduling application in a virtual machine. We also have a couple of Linux VM’s that have various wiki/document management tools that are being evaluated.

I have done some benchmarking of VMWare Server. My benchmarks consisted of Sandra synthetic benchmarks, as well as running SQLIOSIM. I first ran them on a Win2k3 server running native on an HP 385 server. Then I repeated the tests on a Win2k3 VM running on VMWare Server on Linux on the same hardware. My results are showing the VMWare environment to be roughly half as fast. I do believe this is caused by VMWare Server only supporting two cores, when we in fact have two dual core CPUs. This needs to be verified though. Supporting my idea is that the results of benchmarks when VMWare limited the VM to a single CPU were about 1/4 of native speed. Adding the second CPU to the VM doubled the performance.

Our next step is to re-run the benchmarks on a trial version of ESX, but for some reason, the VMWare website is not currently letting me get the trial copy of ESX. With any luck, I will have that problem worked out, and will do the testing later this week.

I have also done some initial testing with the VMWare Converter tool, and will expand on my experiences in another post shortly.

Bookmark to: