LinuxQuestions.org

LinuxQuestions.org (/questions/)
-   Linux - Server (https://www.linuxquestions.org/questions/linux-server-73/)
-   -   Benchmarking Linux servers(Centos 7, SUSE 11, Ubuntu 12.04)-combine optimal stats? (https://www.linuxquestions.org/questions/linux-server-73/benchmarking-linux-servers-centos-7-suse-11-ubuntu-12-04-combine-optimal-stats-4175522753/)

rmcateer 10-20-2014 08:21 AM

Benchmarking Linux servers(Centos 7, SUSE 11, Ubuntu 12.04)-combine optimal stats?
 
I am currently doing a software development Masters project over several months and the project entails benchmarking 3 Linux server versions and making recommendations based on the most optimal one. However, once the optimal one has been chosen, they want me to attempt to write some software which combines the best elements of the three if needed.

They are quite vague on how to do this and after looking online, I was wondering is it even possible. Would it be possible to combine the systems so that, for example's sake, the vmstat of Ubuntu is combined with the iostat of SUSE and the boot time of Centos?

Or is it even possible for me to build my own OS (at a Masters level) using Linux from scratch or something, that will let me create a system that combines the best elements of these three?

Any help or links which will help answer this would be greatly appreciated.

codeguy 10-20-2014 11:46 AM

The types of jobs a server can do are so drastically different, that the phrase "benchmark a server" is completely useless.

"Faster" is never useful. "Faster at xyz" is more useful. A server good with a million http connections is totally different than an oltp database. I'd bet the filesystem you run on has more impact than the OS. For example, a server that hosts very very large video's will run much faster using xfs than ext2.

A server hosting many millions of small files (add/delete/modify) will run much faster on ext4 (probably) than anything else. The difference between OS's would have little effect.

A completely cpu bound job (like compressing a huge file) would probably be exactly the same across any OS and any filesystem.

The differences in default settings between OS's might have an effect. Centos (as a server os), might have some tcp/ip socket settings that Ubuntu (a desktop os) doesn't. So a huge http benchmark could come out different. But, you could make the same changes to Ubuntu and then the benchmarks would come out the same.

I've never tried any of this, its totally opinion.

-Andy

John VV 10-20-2014 03:07 PM

ODD choices
CentOS7.0 just came out the repos are not fully ported yet

SUSE 11 ??
the NON FREE SELS11 sp3 is well NOT FREE
you had to buy a support contract

or
if it is the VERY dead Opensuse 11.1,11.2,11.3,11.4 -- it is unsupported and has been for a long time

ubuntu 12.04
is the OLD long term support
that has been replaced by ubuntu 14.04

RHEL7 and the VERY OLD suse 11 boot VERY differently

it is a "apple VS orange" thing
systemV VS systemD


All times are GMT -5. The time now is 02:10 PM.