Wow.
OpenSolaris PSARC and from Jeff Bonwick’s blog today.
So this really puts into question the value of the Data Domain purchase by EMC. Since most of the Data Domain use cases are as a VTL or NFS backup appliance, I think that once the equivalent Fishworks version hits the street they become direct competition, and the ZFS based solutions are more useful since they can also be used for primary storage.
And for the more adventurous, you can build it yourself for free or pay a reasonable fee for support from Sun.
Wow.
Caveat - have to see how this works in real life, but given the ability to bump up cache memory by using SSDs as L2ARC this means they are a heck of a lot more scaleable than most other solutions (from an architectural standpoint).
Caveat 2 - the ZFS solution is based on straight block calculations from what I gather, whereas the Data Domain solution uses an adaptive block sizing algorithm which could realize higher deduplication rates. But still - “zfs set dedup=on”? It doesn’t get much easier than that.
Update: All of the sudden my auto-replicate script just got a whole lot more efficient with no work on my part. Doing deduplicated snapshot updates across WAN links will be even more efficient. I guess I should add in some good old gzip compression for good measure.