As a ZFS noob I am pretty impressed by the feature set. I'm currently experimenting with it on nexentastor community 3.0.x running as a VM on ESXi4.1 and confirming good dedupe ratios holding Virtual machines on an HP microserver.
From what I understand it's pretty good for data integrity and makes good use of RAM and SSD's as a cache, apart from in one vital area. It can only use 1/4 of your RAM for dedupe metadata as per http://bugs.opensolaris.org/view_bug.do?bug_id=6957289
Currently, I use 4k blocks as the Windows VM's use 4k (aligned) clusters.
Unfortuantely, that makes for a large dedupe overhead or a performance hit as SHA256 hashes are 1/128th of the size of the blocks and these must be stored in RAM.


LinkBack URL
About LinkBacks
Reply With Quote
