SCHED_ULE should not be the default
sgk at troutmask.apl.washington.edu
Thu Dec 15 21:55:55 UTC 2011
On Thu, Dec 15, 2011 at 05:25:51PM +0100, Attilio Rao wrote:
> I basically went through all the e-mail you just sent and identified 4
> real report on which we could work on and summarizied in the attached
> Excel file.
> I'd like that George, Steve, Doug, Andrey and Mike possibly review the
> few datas there and add more, if they want, or make more important
> clarifications in particular about the Xorg presence (or rather not)
> in their workload.
Your summary of my observations appears correct.
I have grabbed an up-to-date /usr/src, built and
installed world, and built and installed a new
kernel on one of the nodes in my cluster. It
CPU: Dual Core AMD Opteron(tm) Processor 280 (2392.65-MHz K8-class CPU)
Origin = "AuthenticAMD" Id = 0x20f12 Family = f Model = 21 Stepping = 2
real memory = 17179869184 (16384 MB)
avail memory = 16269832192 (15516 MB)
FreeBSD/SMP: Multiprocessor System Detected: 4 CPUs
FreeBSD/SMP: 2 package(s) x 2 core(s)
I can perform new tests with both ULE and 4BSD, but you'll
need to be precise in the information you want collected
(and how to collect the data) due to the rather limited
amount of time I currently have.
To summarize my workload, on the master node on my cluster
I start a job that will send N slave jobs to the node of
interest. The slaves perform nearly identical cpu-bound
floating point computations, so the expectation is that
each slave should take nearly the same amount of cpu-time
to complete its task. Communication occurs between only
the master and a slave at the start of the process and
when it finishes. The communication is over GigE ipv4
internal network. The slaves do not read or write to disk.
More information about the freebsd-stable