[Gllug] Why do jobs migrate between cores?

[Nix]

On 27 Jan 2008, Martin A. Brooks verbalised:

> John Winters wrote:
>> 2) Why would a job move from one core to another when there's sod all 
>> else going on?
>
> There's always something else going on.  Even if there's nothing in 
> userland the itself kernel still has many threads running.

Yeah, but they're normally largely idle or blocked unless you're doing
something like USB transfers or software RAID (which does parity
computations in a one-thread-per-array kernel thread).

>                                                          If you want 
> to, you can set a process' affinity so a particular process will always 
> and only run on the designated core.

I *think* you can now use cgroups to constrain processes to arbitrary
sets of CPUs, but I don't actually have access to any multicore systems
so this is all educated guesswork. (My systems are woefully outdated
because I'm so bad at hardware I'm scared to touch them.)

-- 
`The rest is a tale of post and counter-post.' --- Ian Rawlings
                                                   describes USENET
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug