On Tue, Mar 2, 2010 at 9:40 AM, Ric Wheeler <rwheeler@xxxxxxxxxx> wrote: > On 03/02/2010 08:37 AM, Jeff Moyer wrote: >> >> Ric Wheeler<rwheeler@xxxxxxxxxx> 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 -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
