Originally Posted by foljs
It's incredible, but 100% of the comments in this threads are from people that know almost nothing about ZFS, or have only causally used it in some system. Most of them drool over "features" advertised by clueless journalists that know nothing about the actual implementation.
1) Do you know that ZFS (especially snapshots) CONSUMES DISK SPACE LIKE CRAZY?
With over 50% of Mac users on laptops (with only one drive and maybe an external hd), can you really afford to have a hd hungry filesystem?
2) Do you know that ZFS is pretty CPU intensive?
3) Do you know that ZFS needs *A LOT OF MEMORY* and can crash when it runs out?
4) ZFS is expensive for small files.
5) Most of the benefits with pools, unified storage et al means you have to KEEP ALL THE POOL DRIVES connected to use the filesystem, not just plug it when you want it.http://drewthaler.blogspot.com/2007/...ter-redux.html
What the fuck? First off, I take the first paragraph as a huge insult, having used it for quite a while and, even if in small parts, have helped debug it in FreeBSD. I've read white papers, presentations, set up several types of pools, read kernel code, etc etc. I'm don't know "almost nothing" about ZFS. Do YOU
even know what the SPA and DMU layers do?
1) No, I didn't, actually! Unless you delete files like crazy, they don't consume disk space like crazy, since they share all the blocks with the original filesystem until changes occur.
Did YOU know that you can *remove* snapshots if they get too big? It's true!
2) No, not that either. Don't use compression and it's not very CPU intensive at all, at least not when using the default fletcher2 checksum algorithm. Switch to SHA256 and you have only yourself to blame.
3) Again, no. This is mostly/only true on FreeBSD, and you can easily tweak your way around it (and then again, it's mostly/completely solved now on >2GB RAM systems - I'm sure Apple are competent enough to auto-tweak it for systems with less RAM than that).
4) With variable block and stripe sizes? Source, please.
5) No shit, sherlock. How does this differ from other RAID0/RAID5/RAID6 or even JBOD solutions? Use one pool per disk if you for some wacky reason remove them on a regular basis. Or use RAIDZ2 and live with the fact that you've got a semi-degraded array now an then.
Please don't go to personal attacks when YOU are obviously the one with no clue.