I am currently working on a project which involves benchmarking of Linux server systems and making recommendations on that which is best. The project involves running commands such as vmstat on the server versions of SUSE 11, Ubunutu 12.04 and Centos 7. The commands will be managed by bash scripts and a dialog menu box will be made to output the results for user ease.
However, we have now been advised that once the benchmarking results have been investigated and discussed, that we are to build a piece of software that holds the best of these. So for example, if the vmstat statistics are best on Ubunutu and the iostat is best of Centos and the boot time is quickest on SUSE, to put these all onto one system. Is this even possible however? I am not even sure how to build my own Linux OS to store these or if it is possible within the timeframe.
Surely this is not possible given that the syntax is different between the server versions themselves and the only way surely would be to ping the files across to ensure that they are on just one operating system?
Does anyone know if this is possible as I can't seem to see anything online to indicate that it is.
Benchmarking a system as a whole is a complex task, as there are far too many variables involved. The choice of Linux distribution for a company server ends up being a political decision more than a technical one, because the optimal choice depends not only on your benchmarks, but also on maintainer familiarity with the distribution, pre-installed services and the distributions that are already installed in other servers of the company.
You cannot run Ubuntu, Centos and SUSE on the server at the same time. A system with multiple distributions installed will present you with a boot-up screen to choose one of the distros to run. This means that you will not be able to take advantage of Ubuntu's vmstat efficiency, Centos IO optimizations and SUSE boot-up time at the same time, as per your example.
The most informative benchmarks for your clients will be those that imitate your server's purpose in a real-world situation; if this is intended to be a database server then you will have to run the server on Ubuntu with the database for a couple days, then repeat on SUSE, and finally on Centos, or whatever, and design indicators to compare the three sessions on database performance. Not IO, not VM, not boot-up time.