> this is a very good point that I totally overlooked. I concentrated more on > the IOPS alignment plus write durability, > and forgot to check the sequential write bandwidth.
Again, this totally depends on the expected load. Running lots of VMs usually tends to end up being random IOPS on your storage system: Even if all VMs are doing sequential write IO the combined streams would still become random. The bandwidth of fully random IOPS on a spinning disk is usually <1MB... To speed up things you could also use cache tiering http://ceph.com/docs/master/rados/operations/cache-tiering/ This is a new feature in firefly so I've no idea about stability and performance. If you have any control over the usage of the pools you could also consider multiple pools e.g. a SSD only pool for databases and spinners for bulk storage. Cheers, Robert _______________________________________________ ceph-users mailing list ceph-users@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
