How fast is virtualization?

The reason to upgrade to newer equipment is better, faster, smarter. We, me and anyone reading this, know that faster is better. But how to rate the speed of a virtual server?

Simple speed test programs, database speed test, and load testing doesn’t give a good overall impression of the speed of a virtual server.

If you have 10 virtual systems on 1 box is the actual speed equal to 1/10 of the total cpu speed? No, there is overhead taken by the host operating system. So now we have an equation:

virtual system cpu = ( total cpu power - host power ) / number of virtual systems

But with modern virtualization Everything is dynamic. The cpu power floats between systems. Which is great, with a large cluster they can get dynamic systems which float between physical host.

If you have 4 physical quad core hosts, 16 cpu total, how do you rank the total computing power? It isn’t fair to rate it the same way as supercomputing ranking due to each system is running it’s own optimized system for its own purpose, web hosting, database, user management, file and print sharing. But with supercomputer testing, each system is running the same os, same test, and same results.

With some virtual systems, they are high memory access low storage. Others are high bandwidth and low memory requirements.

Is it possible to rank virtual servers? Well the simple way is to put 1 virtual machine and compare the virtual machine system test results to the standard host machine test results.

This is just a start of a discussion to find a way to judge virtualization technology.

No related posts