Another SAMBA Test on Sun Fire X4500 (With fewer disks)
By Tim Thomas on Oct 09, 2007
I received a few emails from the firstname.lastname@example.org alias about the results I posted yesterday...I started to feel curious about what would happen if I had fewer disks in the ZFS storage pool. So today I ran just the RAID-Z test again, but with half the number of disks in the ZFS storage pool shared by SAMBA.
The equipment used was as before except that the Sun Fire X4600 was replaced with a Sun Fire X4200.
The pool was created as below:
zpool create -f sambapool \\
raidz c0t0d0 c1t0d0 c4t0d0 c6t0d0 c7t0d0 \\
raidz c1t1d0 c4t1d0 c5t1d0 c6t1d0 c7t1d0 \\
raidz c0t2d0 c4t2d0 c5t2d0 c6t2d0 c7t2d0 \\
raidz c0t3d0 c1t3d0 c5t3d0 c6t3d0 c7t3d0 \\
spare c0t1d0 c1t2d0 c4t3d0
When I looked at the results, I was at first I very puzzled: the performance (i.e. Total IOPS and Data Rate for the Sun Fire X4500) was identical to the previous test with twice as many disks.
On closer inspection I saw that each individual disk in the storage pool did twice as much work to achieve this vs the earlier test, so the mystery was solved.
Here are the graphs for c0t0d0, for comparison with the previous results: