Here's a good demonstration of why you need to disable the read cache when you're benchmarking a ZFS pool. ;p

The results without disabling it certainly make it seem like this mirror made of a couple old consumer drives can really scream.

#Benchmarking #ZFS

Reply to this note

Please Login to reply.

Discussion

To get a more accurate number, I set the primarycache property of the dataset I was testing to `metadata` temporarily, before returning it to `all`.