On Mon, 2008-02-18 at 10:55 +0100, Peter Zijlstra wrote:
quoted text
> This is my current queue for single RQ group scheduling.

I took these out for a brief maxcpus=1 spin yesterday, and noticed
something.  Running 4 copies of chew-max, one as a user, the context
switch rate was high (~1800/s).  I increased sched_min_granularity_ns to
see if that would lower it, and it did, but it also upset fairness.
With sched_min_granularity_ns bumped to half of sched_latency_ns (40ms
default) it was a largish skew.

top - 06:58:46 up 3 min, 15 users,  load average: 4.71, 2.53, 1.04
Tasks: 214 total,   9 running, 204 sleeping,   0 stopped,   1 zombie
Cpu(s): 39.0%us, 61.0%sy,  0.0%ni,  0.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  P COMMAND
 5389 mikeg     20   0  1464  364  304 R 40.1  0.0   0:29.71 0 chew-max
 5373 root      20   0  1464  364  304 R 19.0  0.0   0:16.30 0 chew-max
 5388 root      20   0  1464  364  304 R 19.0  0.0   0:15.79 0 chew-max
 5374 root      20   0  1464  364  304 R 18.6  0.0   0:15.90 0 chew-max

	-Mike


--
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/