On Thu, Jun 14, 2018 at 11:04 AM Lars Marowsky-Bree <l...@suse.com> wrote:
> Hi all, > > so, I'm wondering right now (with some urgency, ahem) how to make RBD on > EC pools faster without resorting to cache tiering. > > In a replicated pool, we had some success with RBD striping. > > I wonder if it would be possible to align RBD stripe-unit with the EC > chunk size ...? > > Is that worth pursuing at all? How would one best go about this? Yes, you should do this. I think the RADOS code in a recent-ish release fixed things so that it doesn’t allow the EC stripes to be quite as misaligned as before, but it’s still not trying to target anything particularly efficient. I imagine you’d just set it up so that object size divided by k is an even power of two of reasonable size...? > > (I also wonder if the > http://docs.ceph.com/docs/master/rados/operations/erasure-code-profile/ > description of stripe_unit / stripe_width is wrong - shouldn't that be > the number of data+coding chunks, instead of just data chunks? They > need to go somewhere, right?) > > Align k+m and RBD's object_size / (stripe-unit * stripe_count)? Well, if you set an object to 4MB that will be the size of the raw data fed in, right? So if you have 4+2 coding there will be 6 chunks of 1MB each. -Greg > > > Regards, > Lars > > -- > SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB > 21284 (AG Nürnberg) > "Architects should open possibilities and not determine everything." (Ueli > Zbinden) > _______________________________________________ > ceph-users mailing list > ceph-users@lists.ceph.com > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com >
_______________________________________________ ceph-users mailing list ceph-users@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
