Hi, actual status :-). My glusterfs performance is now really great. HW Setup: 2 servers, both with 3 IDE/SATA discs and same disc size 3 x AFR with replicate: 2 over these 2 servers What I did: - wrote a new ext3 filesystem to all of my storage with a bigger inode-size (256byte) - added the io-threads translator on server side - added writebehind, readahead and io-cache translator on client side If I perform a "time dd if=/dev/zeor of=/mnt_glusterfs bs=1M count=10000" on "client A" and parallel a "ls on the mounted gluster share" on "client A" it is still slow, but the reason for this is, that the network traffic from "client A" to the servers has reached the capacity limit (client is writing with 70Mbyte (2x35Mbyte to each gluster server) to the gluster storage (gigabit network)). If I do a "ls" on another glusterfs client (client B) the response is like I do a "ls /"; "ls /home" :-). So, I my "performance problems" are gone. Regards, Matthias Anand Avati schrieb: > And make sure the clients have 'option remote-subvolume' to the volume > name of io-threads and not storage/posix. > > avati > > 2007/8/31, Anand Avati < avati@xxxxxxxxxxxxx > <mailto:avati@xxxxxxxxxxxxx>>: > > One per storage/posix. Cannot have a common io-threads for all. > > avati > > 2007/8/31, Matthias Albert <gluster@xxxxxxxxxxxxxxxx > <mailto:gluster@xxxxxxxxxxxxxxxx> >: > > Hello again :), > > can I add several "storage/posix translators" into one > "performance/io-threads" translator? > Because, I've several discs in one machine, mounted to > different mount > Points. > > So should I add one io-threads translator for one mount Point > (storage/posix translator) or can I add one io-threads volume > for all my > storage/posix volumes on one glusterfsd server? > > Regards, > > Matthias > > > > -- > > It always takes longer than you expect, even when you take into > account Hofstadter's Law. > > -- Hofstadter's Law > > > > > -- > It always takes longer than you expect, even when you take into > account Hofstadter's Law. > > -- Hofstadter's Law
