> Do you think nfs is the problem here? > If so, is there something wrong with our config? > Is it too much to have 10 dir x 10 clients, even if there is almost no traffic? This has very little to do with the number of clients or directories, except that in general terms more users = more work. If the workload is highly random and not easily cached, you could be seeing low throughput but also IOPS starvation. Does 12 disks in Raid6 mean 10 data disks? That's not a lot at all, especially if it is an array that doesn't do much abstraction; any given block will only exist in on a single spindle, creating potential for hot spots if we are talking large sized disks. Can you do any analysis of the array, specifically looking for hot disks, busy disks, high latencies, etc? Do you know what your baseline should be for this array versus your 10 client workload?