On Mon, 2014-02-03 at 17:04 +0100, Flavio Percoco wrote: > On 03/02/14 10:13 -0500, Jay Pipes wrote: > >On Mon, 2014-02-03 at 10:03 +0100, Flavio Percoco wrote: > >> IMHO, the bit that should really be optimized is the selection of the > >> store nodes where the image should be downloaded from. That is, > >> selecting the nearest location from the image locations and this is > >> something that perhaps should happen in glance-api, not nova. > > > >I disagree. The reason is because glance-api does not know where nova > >is. Nova does. > > Nova doesn't know where glance is either. More info is required in > order to finally do something smart here. Not sure what the best > approach is just yet but as mentioned in my previous email I think > focusing on the stores for now is the thing to do. (As you pointed out > bellow too).
Right, which is why I am recommending to get rid of glance-api below... > >I continue to think that the best performance gains will come from > >getting rid of glance-api entirely, putting the block-streaming bits > >into a separate Python library, and having Nova and Cinder pull > >image/volume bits directly from backend storage instead of going through > >the glance middleman. > > This is exactly what we're doing by pulling glance.store into its own > library. I'm working on this myself. We are not completely getting rid > of glance-api but we're working on not depending on it to get the > image data. Cool. Have you pushed a patch for this I can see? Thanks, Flavio! -jay _______________________________________________ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
