[Beowulf] Racing kernels

Robert G. Brown rgb at phy.duke.edu
Tue Apr 5 03:06:17 PDT 2005


On Mon, 4 Apr 2005, Fringe Dweller wrote:

> G'day all.
> 
> Is anyone out there running a "racing" kernel on their nodes?
> 
> By "racing" I mean removing all unnecessary components (SCSI, USB et 
> al), tweaked buffer values, pre-empting, stripped, helium filled, 
> brushed aluminum etc?
> 
> Yes, it's a loaded question. I'm contemplating performance tweaking when 
> I realized that there's a bucket load of stuff in the default kernel 
> config I don't need.
> 
> Now anyone that's ever been exposed to that strange subspecies known as 
> "overclockers" knows you can go to ridiculous lengths ("Duuude, it's 
> fully sick, I can get another 2.3MHz if I take the box outside and hit 
> it with a hose").
> 
> I'd like to think we're giving priority to stability. But still... I 
> think there's scope for "safe" improvement?
> 
> Also need to include the type of target app(s) I s'pose. My Nbody apps 
> are smokin' math stuff but the IO is quite light. So there'd be a fair 
> range of variation in "optimal" config I'd imagine.

The kernel is already pretty modular (depending on how it was built) so
it is likely that only those drivers actually needed are loaded at
runtime.

I tend to think of most kernel bloat as being unused branches outside a
relatively small core -- not a lot saved by trimming them as most of
what the kernel actually does is run the scheduler.  In numerical code
the kernel can be pretty much irrelevant, as long as it isn't openly
inefficient.  Like 99%+ of your CPU goes to the computation, no matter
what you do tweaking the kernel.  Where kernel optimization helps is if
the application USES parts of the kernel a LOT.  Then making small
changes in that branch that make it run more efficient can really
matter.

   rgb

> 
> Thoughts appreciated.
> 
> Cheers
> Steve
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
> 

-- 
Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu





More information about the Beowulf mailing list