Hi people! Happy new year (with old problems...)
Is there some easy or good way to improve cpu load balance on a RedHat enterprise 3 server?

I am running MRTG and some PHP scripts every minute. To distribute the cpu time utilization, I call MRTg and PHP by 12 scripts from second-zero to second-55, taking by division criteria the numbers of network interfaces to monitoring. And I see that there some seconds more heavy, some others more light... but it does not correspond exactly to the number of interfaces I am monitoring.

So, I imagine there are some ways to distribute the job more efficiently and I would like to know if someone has an idea.

thanks Davi

RedHat enterprise 3
that's very old isnt it -- build on linux-2.4.21

the kernel has come a long long way on smp scheduling since then
even newer and better changes usually in -mm kernel version but you know this is one of those things that can never be perfect but they are certainly working on it.

that was rude.
nothing in the Linux kernel is independant of version as changes happen very quickly.
you don't even say if you have a smp or numa system.
what are we suposed to be mind readers ?
and you even put response in red :roll eyes:
it might be (depending on version) that at the top of your kernel source file kernel/sched.c
there are some macro definitions for tweeking.
the only way to mess with scheduling is to change these values and recompile your kernel.
or further alter the kernel source code by changing the algorithms and function called with rebalance_tick() (might be that function depending on what version or what type of system !)