ULE vs. 4BSD in RELENG_7
Josh Carroll
josh.carroll at gmail.com
Fri Nov 2 20:30:36 PDT 2007
> Thank you, that was very useful. I may have something to test very soon.
Sounds great Jeff, just say the word when you need someone to do the
testing. I'll be glad to help!
> What would be interesting to know is if the sum of the temperatures is any
> different. 4BSD gets a much more random distribution of load because a
> thread is run on whatever cpu context switches next. ULE will have
> specific load patterns since it scans lists of cpus in a fixed order to
> assign load.
That seems to correlate with what I'm seeing, since the core that is
"spiking" is cpu0. I will write a little perl script to calculate the
sum periodically and record that. You're right, the one core may be
increasing 1-2 C periodically, but the other cores may remain lower. I
can definitely get you some numbers here.
So I went ahead and ran sysbench to see if changing the
kern.sched.slice value affected that (which I hear is a much more
accurate benchmark of real world MySQL performance than, say,
supersmack). There is a slight hit with the lower slice value, though
it's minimal. Although with 4 threads, the lower slice value actually
increases performance slightly. Here are the results (with 4BSD as a
baseline):
4bsd ule.13 ule.7
2263.6 2250.36 2265.67
2181.18 2310.02 2300.25
2137.87 2304.04 2269.54
2100.41 2252.04 2249.26
buildworld isn't cooperating for me, but once I iron that out, I'll
post some results there as well :)
Thanks once again for all your efforts Jeff. I (and I'm sure others)
really appreciate your work.
Regards,
Josh
More information about the freebsd-performance
mailing list