From: Bernd Eckenfels Date: Sat, 11 Mar 2006 02:57:31 +0100 Subject: [Ocfs2-devel] Ocfs2 performance In-Reply-To: <20060311010913.GN27280@ca-server1.us.oracle.com> References: <20060310002121.GJ27280@ca-server1.us.oracle.com> <20060311010913.GN27280@ca-server1.us.oracle.com> Message-ID: <20060311015731.GA16912@lina.inka.de> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Mark Fasheh Cc: linux-kernel@vger.kernel.org, ocfs2-devel@oss.oracle.com On Fri, Mar 10, 2006 at 05:09:13PM -0800, Mark Fasheh wrote: > until the locks needs to be upgraded or downgraded. This provides a very > large performance increase over always asking the DLM for a new lock. Yes, it is basically the same problem as the buffer cache. Excessive single-use patterns dirty the small cache or require a too big cache to be usefull. Maybe a user specific limit of percentage of hash (and locks) used? I mean the untar test case is a bit synthetic, but think of concurrent read access in a cluster of nntp servers (news article pool). Gruss Bernd