On Fri, 4 Sep 2009, Guiwen wrote: > Hi Sage, > In the ceph website, it said that there would CAS support in the roadmap. Is > this still in the roadmap?
Yeah, although the priority is pretty low. The original idea was the break files into chunks on relatively coarse boundaries (a few MB), and name the chunk object based on the SHA1 (or whatever) hash of the content, with ref counts in the object xattrs. It's conceptually pretty simple. For it to really be a win, though, you need lots of duplicate data in larger files. I'm not convinced the overhead won't be problematic for small chunk sizes... so... it's a lowish priority. > Currently, we use BTRFS as our underline object storage. Do we plan to use > native BTRFS object API to access the objects in BTRFS? It seems we only use > snapshot function from BTRFS. Which object API do you mean? Objects are just stored as files in btrfs. We use the file clone ioctl to dup files (copy-on-write), but not the actual btrfs volume-wide snapshots. The other btrfs bit that's important is an ioctl to hold open a transaction while ceph does a bundle of related operations (usually something like update file1, update file1 xattr, update file2, update file2 xattr, update dir xattr). sage ------------------------------------------------------------------------------ Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july _______________________________________________ Ceph-devel mailing list Ceph-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/ceph-devel
