Oracle to deny deduplication?

Peter Jeremy peterjeremy at acm.org
Fri Sep 24 21:12:07 UTC 2010


On 2010-Sep-22 00:26:20 -0400, Zaphod Beeblebrox <zbeeble at gmail.com> wrote:
>On Tue, Sep 21, 2010 at 3:21 PM, Bob Friesenhahn
><bfriesen at simple.dallas.tx.us> wrote:
>> On Tue, 21 Sep 2010, Zaphod Beeblebrox wrote:
>>>
>>> Shortly thereafter I found a nice table (here:
>>> http://docs.sun.com/app/docs/doc/819-5461/appendixa-1?a=view )
>>>
>>> ... now notice that the official Oracle table lists version 21 of
>>> zpool as "Reserved" ... this worries me.  Is the solution to the
>>> lawsuit against Sun by NetApp (that Oracle settled) that Oracle burys
>>> deduplication?  This could be bad for people who unbury it.
>
>Even if you are correct, I'm still quite wary of Oracle's support of
>ZFS.  Oracle is not known as a "nice" entity.

Agreed, but in this case, I also suspect that it's more of a technical
issue than a legal issue.

>But how would that work --- if you update the pool to 23 (22 or 23 is
>the latest listed there?) then a filesystem with dedup would be read
>... but you might not be able to set the attribute?  Seems odd.

It depends exactly what Oracle have done to the code.  About the only
way to get a definitive answer would be to create a v22 or v23 pool
with dedup enabled and then attempt to import the pool on a S10U9
system.

> The
>ZFS implementation of dedup seems so "sensible" ...

Maybe, but if you read some of the ZFS lists, you'll quickly find that
lots of people ran into problems with it.  Probably the biggest issue
is the massive amount of cache you need with it (either RAM or a
really fast L2ARC) - quite a few people got burnt creating dedup-
enabled pools with snapshots that they then couldn't destroy in a
realistic time.  And my feeling was that there seemed to be a higher-
than-usual level of other issues that people were reporting with dedup.

As a SOHO user, I don't believe I'll be able to justify the cache
costs associated with dedup, though I can see the benefits in some
environments.  Hopefully, the problems will be able to resolved and
dedup will be "officially" available at some point in the future.

-- 
Peter Jeremy
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 196 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-fs/attachments/20100924/a7a19864/attachment.pgp


More information about the freebsd-fs mailing list