From mboxrd@z Thu Jan 1 00:00:00 1970 From: James Bottomley Subject: Re: [Lsf10-pc] [ATTEND] I'd like to attend ;) Date: Thu, 17 Jun 2010 09:02:13 -0500 Message-ID: <1276783333.7285.6.camel@mulgrave.site> References: <4B8C027E.2090709@redhat.com> <4B8C0862.5020308@redhat.com> <4B8D155A.4030508@redhat.com> <4C1A29A5.2070901@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit Cc: Mike Snitzer , linux-fsdevel@vger.kernel.org, Eric Sandeen , Jeff Moyer , lsf10-pc@lists.linuxfoundation.org To: Ric Wheeler Return-path: Received: from bedivere.hansenpartnership.com ([66.63.167.143]:44264 "EHLO bedivere.hansenpartnership.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752566Ab0FQOCQ (ORCPT ); Thu, 17 Jun 2010 10:02:16 -0400 In-Reply-To: <4C1A29A5.2070901@redhat.com> Sender: linux-fsdevel-owner@vger.kernel.org List-ID: On Thu, 2010-06-17 at 09:56 -0400, Ric Wheeler wrote: > On 06/17/2010 09:46 AM, Mike Snitzer wrote: > > On Tue, Mar 2, 2010 at 9:40 AM, Ric Wheeler wrote: > > > >> On 03/02/2010 08:37 AM, Jeff Moyer wrote: > >> > >>> Ric Wheeler writes: > >>> > >>> > >>> > >>>> On 03/01/2010 01:07 PM, Eric Sandeen wrote: > >>>> > >>>> > >>>>> I still need to decide which specific topic to bring/promote, but I'd > >>>>> certainly > >>>>> like to attend, as I have in the past. > >>>>> > >>>>> Generally would like to talk about... > >>>>> > >>>>> Trim/discard plans in filesystem - lately I'm thinking a batch mechanism > >>>>> may > >>>>> be better than trim-as-you-go. > >>>>> > >>>>> Advancements in generic test infrastructures since last year. > >>>>> > >>>>> Could talk a bit about the advancements in proper alignment detection& > >>>>> setup for storage& filesystems but that's not a very big topic. > >>>>> > >>>>> Jan's suggestion of some writeback sanity sounds good to me too. > >>>>> > >>>>> Thanks, > >>>>> -Eric > >>>>> > >>>>> > >>>>> > >>>> I think that all of the above are great topics for this time around. > >>>> > >>>> Specifically, the alignment work has been interesting in that we have > >>>> tried to get not just the kernel bits but the entire stack to work > >>>> properly. > >>>> > >>>> Both this and the discard work should still be quite topical if we can > >>>> get the various storage vendors to evaluate our current upstream bits > >>>> in time :-) > >>>> > >>>> > >>> As you know, I've been picking away at the discard support from the file > >>> system all the way down to the storage. I think I should have some good > >>> numbers and guidance come time for the workshop. > >>> > >>> I am also interested in discussing all things I/O scheduler. > >>> > >>> Oh, and I'd like to attend as well. ;) > >>> > >>> -Jeff > >>> > >>> > >> One topic that I think that might be interesting is to talk about how well > >> we did integrating all of these new features into the stack that cross > >> IO/FS/scheduler boundaries. In a way, we got ahead of the hardware vendors > >> with both the discard support and the various topology bits but did manage > >> to get that supported not just in the kernel but up the tool chain as well. > >> > >> Might be interesting to see where we ended up, think about next steps and > >> if/how we would improve on what we have.... > >> > > I would like attend this year's LSF. > > > > I helped implement and coordinate the development of the I/O Topology > > support through the entire storage stack. My specific contribution > > was in DM and LVM2 but I also worked with Martin Petersen and the > > developers of the layers above the block layer (partition tools, > > filesystems, virtio+qemu). > > > > I'm now looking to implement comprehensive discard support for DM. > > There is a need for various discard support changes in the block layer > > (and below) so it'd be interesting if we touched on this. But my hope > > is that at least some patches will make their way upstream before LSF > > (for 2.6.36 inclusion) -- e.g.: hch's patches to push a discard's > > payload allocation down to the LLD (SCSI or ATA). > > > > I'll also be prepared to discuss aspects of DM if there is any interest. > > > > Mike > > > > I think that these are all great topics - getting discard finished for > DM will be an important goal for the coming year (especially given the > huge surge in SSD devices that really, really can start to use this !!!) Just a procedural note: Since I'm compiling the list of attendee request list, replying to someone else's request is not such a good idea because I went by the thread heads when I did it ... James