From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from psmtp.com (na3sys010amx113.postini.com [74.125.245.113]) by kanga.kvack.org (Postfix) with SMTP id 705756B004A for ; Tue, 3 Apr 2012 16:35:37 -0400 (EDT) Received: from /spool/local by e28smtp02.in.ibm.com with IBM ESMTP SMTP Gateway: Authorized Use Only! Violators will be prosecuted for from ; Wed, 4 Apr 2012 02:05:33 +0530 Received: from d28av02.in.ibm.com (d28av02.in.ibm.com [9.184.220.64]) by d28relay04.in.ibm.com (8.13.8/8.13.8/NCO v10.0) with ESMTP id q33KZGui4300962 for ; Wed, 4 Apr 2012 02:05:17 +0530 Received: from d28av02.in.ibm.com (loopback [127.0.0.1]) by d28av02.in.ibm.com (8.14.4/8.13.1/NCO v10.0 AVout) with ESMTP id q3425qMW007501 for ; Wed, 4 Apr 2012 12:05:53 +1000 Date: Wed, 4 Apr 2012 02:05:00 +0530 From: Srivatsa Vaddagiri Subject: Re: [PATCH 00/39] [RFC] AutoNUMA alpha10 Message-ID: <20120403203500.GA14386@linux.vnet.ibm.com> Reply-To: Srivatsa Vaddagiri References: <1332783986-24195-1-git-send-email-aarcange@redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Disposition: inline In-Reply-To: <1332783986-24195-1-git-send-email-aarcange@redhat.com> Sender: owner-linux-mm@kvack.org List-ID: To: Andrea Arcangeli Cc: linux-kernel@vger.kernel.org, linux-mm@kvack.org, Hillf Danton , Dan Smith , Peter Zijlstra , Linus Torvalds , Andrew Morton , Thomas Gleixner , Ingo Molnar , Paul Turner , Suresh Siddha , Mike Galbraith , "Paul E. McKenney" , Lai Jiangshan , Bharata B Rao , Lee Schermerhorn , Rik van Riel , Johannes Weiner * Andrea Arcangeli [2012-03-26 19:45:47]: > This is the result of the first round of cleanups of the AutoNUMA patch. I happened to test numasched and autonuma against a Java benchmark and here are some results (higher scores are better). Base : 1 (std. dev : 91%) Numa sched : 2.17 (std. dev : 15%) Autonuma : 2.56 (std. dev : 10.7%) Numa sched is ~200% better compared to "base" case. Autonuma is ~18% better compared to numasched. Note the high standard deviation in base case. Also given the differences in base kernel versions for both, this is admittedly not a apple-2-apple comparison. Getting both patches onto common code base would help do that type of comparison! Details: Base = tip (ee415e2) + numasched patches posted on 3/16. qemu-kvm 0.12.1 Numa sched = tip (ee415e2) + numasched patches posted on 3/16. Modified version of qemu-kvm 1.0.50 that creates memsched groups Autonuma = Autonuma alpha10 (SHA1 4596315). qemu-kvm 0.12.1 Machine with 2 Quad-core (w/ HT) Intel Nehalem CPUs. Two NUMA nodes each with 8GB memory. 3 VMs are created: VM1 and VM2, each with 4vcpus, 3GB memory and 1024 cpu.shares. Each of them runs memory hogs to consume total of 2.5 GB total memory (2.5GB memory first written to and then continuously read in a loop) VM3 of 8vcpus, 4GB memory and 2048 cpu.shares. Runs SPECJbb2000 benchmark w/ 8 warehouses (and consuming 2GB heap) Benchmark was repeated 5 times. Each run consisted of launching VM1 first, waiting for it to initialize (wrt memory footprint), launching VM2 next, waiting for it to initialize before launching VM3 and the benchmark inside VM3. At the end of benchmark, all VMs are destroyed and process repeated. - vatsa -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/ Don't email: email@kvack.org