Date: Fri, 25 Nov 2005 07:45:03 +0100 From: Markus Trippelsdorf <markus@trippelsdorf.de> To: Kris Kennaway <kris@obsecurity.org> Cc: freebsd-questions@freebsd.org Subject: Re: so much clock interrupts?! Message-ID: <20051125064503.GA707@bsd.trippelsdorf.de> In-Reply-To: <20051124204359.GD30073@xor.obsecurity.org> References: <20051124013438.T8326@chylonia.3miasto.net> <slrndoarjo.2d8m.Markus@bsd.trippelsdorf.de> <20051124204359.GD30073@xor.obsecurity.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Nov 24, 2005 at 03:43:59PM -0500, Kris Kennaway wrote: > On Thu, Nov 24, 2005 at 07:40:40AM +0000, Markus Trippelsdorf wrote: > > On 2005-11-24, Wojciech Puchar <wojtek@tensor.3miasto.net> wrote: > > > that's what i see in systat with FreeBSD 6.0/i386-SMP > > > > > > > > > 540 534 99 pdwak 2000 cpu0: time > > > 16825 pdpgs 2000 cpu1: time > > > > > > > > > on FreeBSD 5.3 live CD i have 100 not 2000 times per second > > > > > > isn't it too much?! > > > > Yes it is. > > That's why I have kern.hz="100" in my /boot/loader.conf . > > Have you been able to measure a performance benefit from reducing it? > Have you shown that reducing it does not cause your performance to > *drop*? It was increased for a reason..it actually increases > performance on some workloads. > Yes, I guess it increases performance on a high throughput webserver or router that uses polling. But on the Desktop it only increases the overhead without any benefits at all. 2000 interrupts per second per core for the timer is a ridiculous high number and I reduce it simply for aesthetic reasons. This may be a religious issue and everyone should use what he or she seems fitting. -- Markus
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20051125064503.GA707>