On Wed, Dec 03, 2014 at 05:41:31PM -0800, Matthew Ahrens wrote:
> On Wed, Dec 3, 2014 at 5:37 PM, Josef 'Jeff' Sipek via illumos-zfs <
> [email protected]> wrote:
>
> > So, the reason I managed to find this is because I tried to remove the
> > global dbuf hash table - making use of the per-dnode dbuf avl trees to find
> > dbufs.
>
> I imagine that the performance impact of this would be significant.
> Looking up dbufs is in the hot path of cached reads, and is currently
> O(1). Your change would make it O(log(number of dbufs in this dnode)).
Yeah, I'd expect it to really depend on the access pattern. Anyway, it's
relatively easy to rip out the hash and see it suck ;)
Jeff.
--
All parts should go together without forcing. You must remember that the
parts you are reassembling were disassembled by you. Therefore, if you
can’t get them together again, there must be a reason. By all means, do not
use a hammer.
— IBM Manual, 1925
_______________________________________________
developer mailing list
[email protected]
http://lists.open-zfs.org/mailman/listinfo/developer
