Server Fault is a question and answer site for system and network administrators. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am planning on building a Continuous Integration Server (the existing one is too slow). My plan is to run multiple OSes for our application testing on KVM.

The structure will be a single physical machine with 6 CPU cores and 3 hard drives, running 3 KVM instances at a time.

I plan to run each OS with 2GB of RAM, 2 cores, and from a unique disk (every disk having a copy of all images).

Can I get any benefit from this architecture? Especially IO? How much difference in IO performance is there between having VM image vs a physical system running with 2 cores and 2GB of RAM? Will be dedicated disk make any noticeable difference?

share|improve this question
Depends on your OS and the type of machine you are virtualizing. What is at the moment "too slow"? Are there any heavy write to disks involved? – Nils Jul 31 '11 at 20:08

The amount of I/O performance you can get from 6 six disks is the same regardless of whether they're carved up into three RAID1 volumes, or one big RAID10. The advantage of splitting it up into three is that the VMs on those volumes can't trigger I/O starvation for the other VMs. The advantage to having one big volume is that occasional peak loads on one VM are much less likely to trigger I/O starvation as the overhead available should be a lot larger. That said, if all of your VMs will be running the same distributed workload and are likely to be busy at the same times, there is no clear advantage to either method.

As for the performance of physical vs. virtual, I don't know what the performance multiplier is for KVM/QEMU these days.

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.