On Apr 14, 2010, at 11:10 PM, Daniel Carosone wrote:
> On Wed, Apr 14, 2010 at 09:58:50AM -0700, Richard Elling wrote:
>> On Apr 14, 2010, at 8:57 AM, Yariv Graf wrote:
>>
>>> From my experience dealing with > 4TB you stop writing after 80% of zpool
>>> utilization
>>
>> YMMV. I have routinely completely filled zpools. There have been some
>> improvements in performance of allocations when free space gets low in
>> the past 6-9 months, so later releases are more efficient.
>
> Some weeks ago, I read with interest an excellent discussion of
> changes resulting in performance benefits for the fishworks platform,
> from Roch Bourbonnais.
>
> After all the analysis, three key changes are described in the
> penultimate paragraph. The first two of these basically adjust
> thresholds for existing behavioual changes (e.g the switch from
> first-fit to best-fit); the last is an actual code change.
>
> I meant to ask at the time, and never followed up to do so, whether:
> - these changes are also/yet in onnv-gate zfs
> - which builds, if so
> - whether the altered thresholds are accessible as tunables, for
> older builds/in the meantime.
There are several
b114: 6596237 Stop looking and start ganging
b129: 6869229 zfs should switch to shiny new metaslabs more frequently
b138: 6917066 zfs block picking can be improved
there are probably a few more...
-- richard
> I've just added the above as a comment on the blog post, in the
> hopes of attracting Roch's attention there. There have been recent
> commits go by (>b134) that seem promising too.
>
> --
> Dan.
ZFS storage and performance consulting at http://www.RichardElling.com
ZFS training on deduplication, NexentaStor, and NAS performance
Las Vegas, April 29-30, 2010 http://nexenta-vegas.eventbrite.com
_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
