Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756175AbYJVNFw (ORCPT ); Wed, 22 Oct 2008 09:05:52 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752059AbYJVNFp (ORCPT ); Wed, 22 Oct 2008 09:05:45 -0400 Received: from mail.gmx.net ([213.165.64.20]:44385 "HELO mail.gmx.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with SMTP id S1752043AbYJVNFo (ORCPT ); Wed, 22 Oct 2008 09:05:44 -0400 X-Authenticated: #14349625 X-Provags-ID: V01U2FsdGVkX18A7Bl+PPMAOE5Pdqa0X9xyIndkt2B40TOFmvk+Sg cFepfHcVClZkS9 Subject: Re: [PATCH 0/4] pending scheduler updates From: Mike Galbraith To: Ingo Molnar Cc: Srivatsa Vaddagiri , Peter Zijlstra , LKML In-Reply-To: <20081022124240.GD25536@elte.hu> References: <20081017172701.047939625@chello.nl> <20081020120538.GD10594@elte.hu> <20081021173529.GG11679@linux.vnet.ibm.com> <20081022094019.GG12453@elte.hu> <1224669821.6871.16.camel@marge.simson.net> <1224671528.7511.11.camel@marge.simson.net> <20081022121007.GG8095@elte.hu> <1224679082.12294.15.camel@marge.simson.net> <20081022124240.GD25536@elte.hu> Content-Type: text/plain; charset=utf-8 Date: Wed, 22 Oct 2008 15:05:12 +0200 Message-Id: <1224680712.15904.21.camel@marge.simson.net> Mime-Version: 1.0 X-Mailer: Evolution 2.22.1.1 Content-Transfer-Encoding: 8bit X-Y-GMX-Trusted: 0 X-FuHaFi: 0.65 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 1117 Lines: 29 On Wed, 2008-10-22 at 14:42 +0200, Ingo Molnar wrote: > * Mike Galbraith wrote: > > > > > With that patch, pgsql+oltp scales perfectly. > > > > > > hm, tempting. > > > > I disagree. Postgres's scaling problem is trivially corrected by > > twiddling knobs (or whatnot). [...] > > okay, then we need to document it a bit more: what knobs need twiddling > to make it scale perfectly? Twiddle sched_wakeup_granularity_ns or just turn FAIR_SLEEPERS off for best pgsql+oltp scalability. Pgsql+oltp collapse is delicate. It doesn't take much to send it one way or the other. Cut preempt a wee bit, and it can snap right into shape. I think we need to penalize sleepers a wee bit as load climbs in general though, everybody begins to like preemption less as load increases. Mysql+oltp improves as well, and it loves preempt at modest load. -Mike -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/