November 28, 2006

RAID5 initial benchmarks

Now that its been up a few weeks, the hardware RAID5 card seems to be running well.

Whew.

Today, with an idling system, I did some simplest block-IO benchmarks. This just writes and then reads a large number of 1B blocks to and from disk. There’s 1GB of memory in the system, so this should well exceed the IO kernel cache.

time dd if=/dev/zero of=1G bs=1M count=1024
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB) copied, 17.27 seconds, 62.2 MB/s

real    0m17.283s
user    0m0.000s
sys     0m2.720s
usul:/mnt# sync
usul:/mnt# time dd if=1G of=/dev/null bs=1M
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB) copied, 25.9455 seconds, 41.4 MB/s

real    0m25.947s
user    0m0.004s
sys     0m2.700s
usul:/mnt# time dd if=/dev/zero of=2G bs=1M count=2048
2048+0 records in
2048+0 records out
2147483648 bytes (2.1 GB) copied, 46.3935 seconds, 46.3 MB/s

real    0m46.410s
user    0m0.016s
sys     0m5.428s
usul:/mnt# time dd if=2G of=/dev/null bs=1M
2048+0 records in
2048+0 records out
2147483648 bytes (2.1 GB) copied, 55.5036 seconds, 38.7 MB/s

real    0m55.573s
user    0m0.008s
sys     0m5.480s
usul:/mnt# time dd if=2G of=/dev/null bs=1M
2048+0 records in
2048+0 records out
2147483648 bytes (2.1 GB) copied, 56.2326 seconds, 38.2 MB/s

real    0m56.294s
user    0m0.020s
sys     0m5.624s

Not bad at all, not fibre channel but respectable. I expected a bit faster, so I’ll have to dig a bit and see what’s going on. OK for now, though.


Debian Linux RAID


Previous post
Comfort food and gender Brian Wansink, of the Cornell Food and Brand Lab, has a new book out, and an interview on Salon. Most of it I knew, but this piece of insight caught
Next post
Another watch project If only I had my own CNC, I too could find a WW2-vintage Hamilton pocketwatch and recase it as a wristwatch. Milling the case out of solid