Virtualbox, consistent performance with numerous guests?
I have a quad-core computer and I'm running Linux amd64 as host OS.
I want to run 3 guest Linux systems in VirtualBox, all with only one allowed CPU core, and 80% CPU cap, maybe I'll increase cap, but not to 100%. (I'm saving one core for the host os).
All these guests are exactly same configuration, same distribution as guest os and same applications.
I'd like to do some testing with 1, 2, 3 running guests. Meanwhile, I'm trying to minimize host os workload by killing everything that's not needed.
The question is, how consistent performance will I get with my guests? I'm thinking of jumping from one core to another (this applies some overhead I guess), etc. Does it makes any sense doing performance-related tests like this? If the answer is no, do you have any advices for me? Some other solution than Virtualbox, some configuration tips, etc?
ps.: I'm currently not allowed to buy or borrow dedicated hardware. I'm trying virtualbox because (maybe!) I can run totally equal nodes for my test, and I could extrapolate the result.
I have a quad-core computer and I'm running Linux amd64 as host OS.
I want to run 3 guest Linux systems in VirtualBox, all with only one allowed CPU core, and 80% CPU cap, maybe I'll increase cap, but not to 100%. (I'm saving one core for the host os).
All these guests are exactly same configuration, same distribution as guest os and same applications.
I'd like to do some testing with 1, 2, 3 running guests. Meanwhile, I'm trying to minimize host os workload by killing everything that's not needed.
The question is, how consistent performance will I get with my guests? I'm thinking of jumping from one core to another (this applies some overhead I guess), etc. Does it makes any sense doing performance-related tests like this? If the answer is no, do you have any advices for me? Some other solution than Virtualbox, some configuration tips, etc?
ps.: I'm currently not allowed to buy or borrow dedicated hardware. I'm trying virtualbox because (maybe!) I can run totally equal nodes for my test, and I could extrapolate the result.
No comments:
Post a Comment