On Tue, 8 Jun 2010, Tomasz Chmielewski wrote: > I'd like to build a cluster for storing large amounts of data. > > The cluster would consist of slow (cheap) storage and fast (expensive) > storage. > > I would like to migrate data from fast (expensive) storage to slow (cheap) > storage if it wasn't accessed for, say, 6 months. > > Would ceph somehow help me achieve that? We don't currently have that functionality. In principle it's certainly something the system could do, and the underlying storage architecture lends itself to that: you can create different object storage pools, with different data distribution (crush) rules, mapped to different devices, etc. What's missing is the HSM process that goes and finds old, idle data and migrates it. And atime tracking. You can also use a mix of small+fast and big+slow disks/ssds with a single pool of storage by writing a crush rule that puts the primary replica on the fast disks and additional replicas on slow disks. (The primary replicas handle the read workload.) That approach fixes the ratio between fast and slow capacity, though. You could also build each osd volume with a mix of fast and slow disks/ssds, and use btrfs or some other local fs to manage the distribution of objects across disk types. (Btrfs doesn't do that yet, but presumably will at some point.) sage -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
