Return-Path: linux-nfs-owner@vger.kernel.org Received: from mail-we0-f177.google.com ([74.125.82.177]:52335 "EHLO mail-we0-f177.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751143Ab3I2Lyr (ORCPT ); Sun, 29 Sep 2013 07:54:47 -0400 Received: by mail-we0-f177.google.com with SMTP id t60so4238648wes.8 for ; Sun, 29 Sep 2013 04:54:45 -0700 (PDT) Message-ID: <52481502.1010108@primarydata.com> Date: Sun, 29 Sep 2013 14:54:42 +0300 From: Benny Halevy MIME-Version: 1.0 To: Christoph Hellwig CC: "J. Bruce Fields" , NFS list , Lev Solomonov , Idan Kedar , Nadav Shemer Subject: Re: [PATCH RFC v0 0/49] pnfsd-dlm References: <52447EA0.7070004@primarydata.com> <20130929114240.GA25750@infradead.org> In-Reply-To: <20130929114240.GA25750@infradead.org> Content-Type: text/plain; charset=ISO-8859-1 Sender: linux-nfs-owner@vger.kernel.org List-ID: On 2013-09-29 14:42, Christoph Hellwig wrote: > On Thu, Sep 26, 2013 at 02:36:16PM -0400, Benny Halevy wrote: >> The following patchset implements an extension to nfsd >> providing a complete minimal pnfs server exporting >> DLM-based clustered file systems such as GFS2 or OCFS2. > > Does this actually buy us anything by now? Last time I saw numbers for > this implementation it was slower than an active/passive setup over > those filesystem due to the way their cluster locking works. > Was this for write sharing or read only? As far as I understand, the share locks do not cause arbitration and they do provide bandwidth scalability via multiple nodes. That said, it is worth measuring. Benny