Dedup doesn''t really induce extra CPU usage - you need to calculate
checksums anyway. However it does require sha256 checksums instead of
fletcher but still, usually it''s not CPU which is a problem.
But then being able to dedup your data later could make your writes much
faster depending on configuration and workload.
This should be possible once the famous bp_rewriter hits public.
--
Robert Milkowski
http://milek.blogspot.com
> -----Original Message-----
> From: zfs-code-bounces at opensolaris.org [mailto:zfs-code-
> bounces at opensolaris.org] On Behalf Of Siegfried Leonard
> Sent: 02 April 2011 10:08
> To: zfs-code at opensolaris.org
> Subject: [zfs-code] Manual deduplication scrubs
>
> Hello all,
>
> Is it possible to do deduplication in place when the system is idle? That
> would be pretty useful to keep performance on low memory systems or not
> increase CPU usage of slower machines such that the filesystem
> performance would be decreased if using the live approach.
>
> Is this even feasible? Would the dedup database be updated from the last
> run? Kind of like a ZFS snapshot being taken, except in this case I
don''t
think> it can be implemented using ZFS snapshots, or can it? Would the
> performance benefit of this be worth enough implementing?
>
> Just some thoughts, ideas, and questions. Let me know your opinion.
>
>
> Cheers,
> Siegfried
> --
> This message posted from opensolaris.org
> _______________________________________________
> zfs-code mailing list
> zfs-code at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/zfs-code