HZ and HPC
Benjamin C.R. LaHaise
bcrl at kvack.org
Wed Jan 29 11:50:57 PST 2003
On Wed, 29 Jan 2003, Ken Chase wrote:
> If anyone can point out to me how to achieve something like HZ ~ 10 such that
> cache thrashing is a non issue, I'd love to hear - need some help. I'd think
> that HZ ~ 10 would lead to some very interesting interupt handling problems
> for devices with little or no buffer, among other things. Defintely a non
> trivial kernel design problem -- need to avoid blocking on priority
> inversion among other things.
Under Linux, the core kernel supports running at arbitrary HZ settings in
2.5. For 2.4, there are patches needed as a few places expose HZ directly
to userspace and that results in minor breakage of the ABI. IBM has done
some work on s/390 and ARM (not sure on that one) for the mobile Linux
watch to only enable timer ticks when there is a pending timer request.
This results in significant power savings for the watch, and lots of CPU
savings on mainframes running oodles of virtual machines.
Iow, yes, people are aware of the problem and do want to fix it, but the
patches may not be directly useable for you on x86 at present.
More information about the Beowulf