vmware vs normal server

Question

what is a good benchmark for me to run on my old server and my new vmware server? I want to verify that I have not lost any performance between the two

"want to verify that I have not lost any performance" -- pretty sure it's not possible to not lose *any* performance. — Swoogan, Dec 15 '09 at 18:07
@Swoogan, it's entirely possible to *gain* performance if the new hardware is significantly better than the old hardware. — ThatGraemeGuy, Dec 15 '09 at 19:40
We could do with some clarification here. Is the virtual machine now running on the same hardware as the old machine? If so you will most definitely lose at least some performance. If not, on what basis do you what to make the comparison? — John Gardeniers, Jan 27 '10 at 00:45

score 1 · Answer 1 · answered Dec 15 '09 at 18:06

I want to verify that I have not lost any performance between the two

Since you're running VMWare but didn't specify the version, I'd say start with VMWare VMMark. It's free, but I'm not sure if your specific setup meets the minimum requirements in order to run it.

What do you want to focus in on in terms of performance? CPU? RAM? Disk I/O? Network I/O? VMGuest performance?

I'd like to save you the time and effort (if possible) so I'm sorry to say, but in virtualization you will lose a little bit of performance by default depending on the solution VMWare, Xen, QEMU, Virtual Iron, etc. etc. It's a trade-off between bare-bones performance (1 OS to 1 server) and virtualization emulation/translation/hardware assist (N+ OSes to 1 server), etc. However, losing 3-4% here or there overall is still worth the savings if performance isn't priority #1.

Note through that virtualization will bring out the worst of your server fast. Many people overlook that 10 vm's means 10 tims the disc IO (simplified) and a normal disc subsystem (one or 2 discs) will overload. plan accordingly ;) — TomTom, May 07 '10 at 09:06

score 0 · Answer 2 · answered Dec 15 '09 at 15:13

try to run any appropriate benchmarks on your VM while doing some heavy loading on the same resource on a different VM. Usually the most critical are disk and network access.

for example, if you want to benchmark disk access, first do a 'low contention' test, then start some heavy 'disk trashing' (maybe untar ten big packages simultaneously) process on a different VM while repeating the benchmark.

if you want to benchmark network access, do the test while at the same time moving lots of data in and out from a different VM to the 'external' world.

in short, you have to be sure that you'll get reasonable performance no matter how badly is somebody else fighting for the same physical resources.

(and yes, this should be on ServerFault)

vmware vs normal server

2 Answers2