From mboxrd@z Thu Jan 1 00:00:00 1970 From: Joel Becker Date: Mon, 5 Apr 2010 18:36:03 -0700 Subject: [Ocfs2-devel] [PATCH 0/4] Ocfs2 allocation tuning In-Reply-To: <1270516636-12180-1-git-send-email-mfasheh@suse.com> References: <1270516636-12180-1-git-send-email-mfasheh@suse.com> Message-ID: <20100406013602.GA15139@mail.oracle.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: ocfs2-devel@oss.oracle.com On Mon, Apr 05, 2010 at 06:17:12PM -0700, Mark Fasheh wrote: > The following patches apply some lessons I've learned during some testing of > a couple non-trivial workloads on Ocfs2. Roughly speaking, they tune "up" > the default local alloc windowxs, and tune "down" the reservations code a > bit. > > I primarily ran two tests - one simulating a large multi-node data-processing > workload, and the multi-threaded writers test, mtwrite at: > > http://www.kernel.org/pub/linux/kernel/people/mfasheh/ocfs2/tests/mtwrite.c > > All patches were developed and tested on top of the current set of > allocation changes in ocfs2.git (primarily reservations, and the local alloc > fixes we have). A git branch containing all the patches is available, based > off 2.6.33: > > git pull git://git.kernel.org/pub/scm/linux/kernel/git/ocfs2-mark.git disk-alloc > > Basically though, this all involved many many test runs (of which I have > lots of data) and a thorough checking of fragmentation levels. I changed > only one value at a time. Details regarding some of the test runs are within > the specific patch descriptions. These patches are now in the merge-window branch of ocfs2.git. Joel -- "It is not the function of our government to keep the citizen from falling into error; it is the function of the citizen to keep the government from falling into error." - Robert H. Jackson Joel Becker Principal Software Developer Oracle E-mail: joel.becker at oracle.com Phone: (650) 506-8127