<div dir="ltr"><div>Nice work.</div><div><br></div>Looks like FreeBSD 10 is not fully utilizing all the cores or there are some serious migration/contention issues going on. Trying FreeBSD 9.x might help in pointing to any 10.x (still a young distro) issues. You may also want to try some artificial benchmarks that are used specifically for testing SMP scaling - like the NASA NPBs - <a href="https://www.nas.nasa.gov/publications/npb.html">https://www.nas.nasa.gov/publications/npb.html</a> .<div>
<br></div><div>Brett</div></div><div class="gmail_extra"><br><br><div class="gmail_quote">On Mon, Mar 10, 2014 at 4:58 AM, Francois Tigeot <span dir="ltr"><<a href="mailto:ftigeot@wolfpond.org" target="_blank">ftigeot@wolfpond.org</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="">On Mon, Mar 10, 2014 at 09:46:25AM +0100, Sascha Wildner wrote:<br>
><br>
> The FreeBSD numbers certainly don't look right (as in, realistic). Are you<br>
> sure there is no configuration issue?<br>
<br>
</div>They are 100% reproductible; I installed FreeBSD on both ZFS and UFS2 to be<br>
sure ZFS didn't play any trick with memory but to no avail.<br>
<br>
I didn't use any special configuration apart from the Postgres changes. If<br>
there's some magical switch to make FreeBSD perform better, I'd like to know<br>
it.<br>
<span class="HOEnZb"><font color="#888888"><br>
--<br>
Francois Tigeot<br>
</font></span></blockquote></div><br></div>