Date: Mon, 14 Jun 2004 02:56:13 -0500 From: Jon Noack <noackjr@alumni.rice.edu> To: Luigi Rizzo <rizzo@icir.org> Cc: current@freebsd.org Subject: Re: Device polling Message-ID: <40CD5A1D.2070105@alumni.rice.edu> In-Reply-To: <20040614002429.A16826@xorpc.icir.org> References: <Pine.BSF.4.58L0.0406110957020.49590@crimp.gluerecord.net> <40CC97E0.5010003@alumni.rice.edu> <20040613181315.GB54104@ip.net.ua> <40CCC201.6080804@alumni.rice.edu> <20040613223658.GB55275@ip.net.ua> <20040614002429.A16826@xorpc.icir.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On 06/14/04 02:24, Luigi Rizzo wrote: > On Mon, Jun 14, 2004 at 01:36:58AM +0300, Ruslan Ermilov wrote: >> On Sun, Jun 13, 2004 at 04:07:13PM -0500, Jon Noack wrote: >>> On 06/13/04 13:13, Ruslan Ermilov wrote: >>>> On Sun, Jun 13, 2004 at 01:07:28PM -0500, Jon Noack wrote: >>>>> I just tested this on my SMP all-in-one home server (Web, >>>>> Mail, > ... >> Bump HZ. If you were using the 64-bit PCI I'd say bump it to 5000, >> but since you're using 32-bit PCI, bumping it that high just >> doesn't make sense, so raise it up to 2000-3000. >> >> Set kern.polling.idle_poll=1. > ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ > this is the part that cannot possibly work with SMP, > and that seems to have problem with current even on !SMP Well, in limited testing it doesn't seem to have any effect. After a lot of tweaking, polling on my SMP box appears to lower throughput by ~50Mbps (~370 down to ~320). It does improve performance on my single-processor machines, though. Again, if I have more time I'll give the netperf patches a shot. Jon
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?40CD5A1D.2070105>