Re: Storage Claus on Partitioned Global Index
From: UXDBA <unixdba73_at_googlemail.com>
Date: Wed, 18 Nov 2009 07:07:30 -0800 (PST)
Message-ID: <e67ce53c-bde8-4a72-a405-80fe52725e2b_at_b15g2000yqd.googlegroups.com>
On Nov 18, 11:53 am, Noons <wizofo..._at_yahoo.com.au> wrote:
> Steve Howard wrote,on my timestamp of 18/11/2009 1:38 PM:
>
> > This would make sense for the original partitions at creation, but I
> > don't understand why a split (which effectively adds a new partition
> > P4 in my example without a tablespace clause) doesn't go into the
> > "default" USERS tablespace (whatever that means).
>
> The split is from an existing partition, which is in a given tablespace. The
> result should go into same tablespace, in the absence of any instructions saying
> otherwise. It "inherits" defaults from the original partition.
> Or at least that's my reading of the doco and what I've seen so far.
>
> >> "Any attributes you do not specify are inherited from the original
> >> partition." Of course, that is talking about tables, so the docs
> >> ought to mention the same thing about indices.
>
> Hmmm, I've found that when it comes to partitions, indexes tend to "behave" very
> much as tables do - as far as syntax goes, of course.
>
> > I'm not sure what "original partition" means in that case? What if I
> > created two partitions in two different tablespaces "originally"?
>
> At a guess and without trying:
>
> partition 1, in tablespace one, split, two resulting partitions end up in
> tablespace one.
>
> partition 2, in tablespace two, split, two resulting partitions end up in
> tablespace two.
>
> Assuming of course defaults.
>
> Now: if someone can show me how to make sure partitions compress the same as a
> non-partitioned table...
> I've just hit that one a few weeks ago, with 10.2.0.3:
>
> Non-partitioned table, compressed: smaller size than original.
> Partitioned with compression: same total size as original uncompressed.
> Meh!...
Date: Wed, 18 Nov 2009 07:07:30 -0800 (PST)
Message-ID: <e67ce53c-bde8-4a72-a405-80fe52725e2b_at_b15g2000yqd.googlegroups.com>
On Nov 18, 11:53 am, Noons <wizofo..._at_yahoo.com.au> wrote:
> Steve Howard wrote,on my timestamp of 18/11/2009 1:38 PM:
>
> > This would make sense for the original partitions at creation, but I
> > don't understand why a split (which effectively adds a new partition
> > P4 in my example without a tablespace clause) doesn't go into the
> > "default" USERS tablespace (whatever that means).
>
> The split is from an existing partition, which is in a given tablespace. The
> result should go into same tablespace, in the absence of any instructions saying
> otherwise. It "inherits" defaults from the original partition.
> Or at least that's my reading of the doco and what I've seen so far.
>
> >> "Any attributes you do not specify are inherited from the original
> >> partition." Of course, that is talking about tables, so the docs
> >> ought to mention the same thing about indices.
>
> Hmmm, I've found that when it comes to partitions, indexes tend to "behave" very
> much as tables do - as far as syntax goes, of course.
>
> > I'm not sure what "original partition" means in that case? What if I
> > created two partitions in two different tablespaces "originally"?
>
> At a guess and without trying:
>
> partition 1, in tablespace one, split, two resulting partitions end up in
> tablespace one.
>
> partition 2, in tablespace two, split, two resulting partitions end up in
> tablespace two.
>
> Assuming of course defaults.
>
> Now: if someone can show me how to make sure partitions compress the same as a
> non-partitioned table...
> I've just hit that one a few weeks ago, with 10.2.0.3:
>
> Non-partitioned table, compressed: smaller size than original.
> Partitioned with compression: same total size as original uncompressed.
> Meh!...
thanks all for your investigations. Received on Wed Nov 18 2009 - 09:07:30 CST