Unix & Linux Stack Exchange is a question and answer site for users of Linux, FreeBSD and other Un*x-like operating systems. 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 have several long running benchmarks (SPEC CPU 2006 benchmarks) to run on a Linux server.

The server is running Gentoo Linux with a Linux Kernel 3.6.11.

I saw some big differences between different runs and I'm thinking that it could be some problem with my configuration.

I'm the only user of the server.

I have already done several configuration changes:

  • I have disabled the CPU frequency scaling feature of the kernel.
  • I have removed all cron jobs
  • I made sure that there were no frequency scaling in the BIOS

Is there some others things I should configure/disable/enable in the kernel or on the installation ?

I do not need to make the server faster, but I want to make sure that the benchmarks are not encountering pertubations


share|improve this question
  1. Disconnect from network

    Do you have physical access to the server? Is it possible to disconnect the sever from network while running the test? Serving network request(eg. http/sql) during test will very likely affect result.

  2. Disable cron job

    Cron job during test will also affect test result, especially heavy one like indexing or backup routine.

share|improve this answer
I have physical access, but I would prefer no disconnecting the network so that I can run benchmarks from home too. There are no accesses to the server, expect broadcasting from the network. Do you think that is enough to cause variations ? I will disable cron jobs but I don't think there is any. – Baptiste Wicht Jan 20 '13 at 23:54
how much discrepancy you are seeing? Maybe posting the result will help pinpoint the reason. – John Siu Jan 20 '13 at 23:56
@JohnSiu is right. Some discrepancy is to be expected. If there is too much discrepancy, perhaps the benchmark itself is flawed (doing something that gets cached the first time around, and you see just lookup costs later, for example). – vonbrand Jan 21 '13 at 17:25
I'm seeing sometimes up to 30% differences in the scores of the benchmarks between two runs. Normally, this benchmark suite is known to be very stable and very well done. – Baptiste Wicht Jan 21 '13 at 20:00
The SPEC test suits are not free so I cannot test them. How long the benchmark takes per run? Is it practical to do 5 or more runs from a script and see if there is a trend? – John Siu Jan 21 '13 at 20:10

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.