I found a neat way of benchmarking filesystems this morning. Using the following command:
dd bs=1000000 count=1000 if=/dev/zero of=testfile & speedometer testfile

Basically it monitors the creation of the file and plots a graph as it’s getting created. Running it on my ZFS partition (with dedup and compression), it got about 10MB/s, and on my RAID5 it got 100MB/s. It’s rather surprising because one would think that ZFS could write a whole bunch of zeros really easily. Clearly not. I’m beginning to form a strong dislike of ZFS.

András Fekete