At Fri, 07 Oct 2011 10:28:34 -0700, Mark Pace wrote: > > > On 10/7/2011 1:20 AM, MORITA Kazutaka wrote: > > At Thu, 06 Oct 2011 23:16:09 -0700, > > Mark Pace wrote: > >> Thanks for the info on the collie commands. > >> > >> I think the key advantage to having a "canonical" backup store would be that you didn't have to stress the collie cluster with backup requests because the data would already be on that one unit. With the data multicasting around, it seems a shame to run up the cluster for a redundant request if we could create the node that stores it all by default. > > But the backup node limits the total storage size of Sheepdog, doesn't > > it? I think we cannot create a large cluster with the architecture. > Yes, I do believe the backup node would limit the size of the cluster as > it alone would have to have as much storage as the total provided by the > cluster. I wasn't worried about having to have a backup node have some > serious amounts of storage, but it could be an issue. Other than the storage size issue, the backup node would be a bottleneck if there are many VMs. The backup node requires a huge amount of disk space and bandwidth, but if we could use such machine, we wouldn't need a clustered storage system. However, on a small cluster environment with a few nodes, the backup node idea looks good. If someone sends a patch to support it, I'll accept it. :) Thanks, Kazutaka |