Thanks to all of you who replied.

Various people pointed me at statit as the particularly useful tool for spotting what's going on. I have two volume groupings, one with 3 disks, and one with 4. My data is only on one of those volume groups, and statit clearly shows the usage on the relevant disks being very high, up to the point where I'm thrashing the cache on the filer.

So I'll split my data across both volume groups, which should give me better throughput.

Thanks again,

Edward.

Disk statistics (over 31 seconds) ut% is the utilization % of the disk; chain is avg number of 4K blocks/op; usecs is round trip time to disk in microseconds per 4K block.

disk ut% xfers/s urds/s chain usecs cwrs/s chain usecs crds/s chain usecs 8.0 93 171.12 118.70 1.06 5711 26.90 3.45 2213 25.51

1.70

2592 8.1 90 168.80 114.54 1.04 5843 28.09 3.16 2433 26.16

1.71

3163 8.2 0 0.38 0.00 0.00 0 0.38 3.16 1605 0.00

0.00

0 8.3 28 79.06 0.25 1.00 9500 38.35 4.08 1340 40.45

1.73

1112 8.4 90 170.70 118.45 1.05 5771 27.80 3.11 2359 24.45

1.69

3009 8.5 0 0.67 0.19 1.00 3833 0.19 3.83 1565 0.29

1.22

1545 8.6 0 0.61 0.12 1.00 7750 0.35 2.81 1580 0.12

1.75

1428