> Hmm, why? An nfs client will do just what you want in terms of assigning a > single address for the connection.
In my experience, Linux NFS is slow, especially for writes. I know, when I say that, you think I probably do not know how to tune my network, storage array, or NFS settings. You think maybe I am using UDP or small packets or "sync" (client or server) or... And that is OK, because I would probably think the same thing if our positions were reversed. But I am able to get better than 1 gigaBYTE per second sustained NFS read speeds (sustained = hundreds of gigabytes) with a cold cache, between system A and system B in either direction. I can get better than 1 gigabyte per second sustained local writes on either system. But I just cannot get NFS writes to go faster than 200 megabtyes per second or so. (I can do an FTP "put" at 1 gigabyte per second.) I am hoping the native gluster protocol can outperform this. > The Gluster client, by contrast, > downloads the specification of the storage cluster on initially establishing > the connection, and works from that. Does that configuration use host names, or IP addresses? (Yes, what I am considering is a kludge...) Thank you for your responses, by the way. - Pat _______________________________________________ Gluster-users mailing list [email protected] http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
