Comments on: Lunchtime Quiz

By: Struttura interna di un blocco dati Oracle « Oracle and other

Struttura interna di un blocco dati Oracle « Oracle and other — Wed, 10 Jun 2009 10:03:29 +0000

[...] po’ di tempo fa (esattamente era il 20 maggio) Jonathan Lewis ha lanciato sul suo blog un quiz che chiedeva per una tabella (di cui riportava lo script di creazione) con una colonna e con il [...]

By: Log Buffer #147: a Carnival of the Vanities for DBAs | Pythian Group Blog

Log Buffer #147: a Carnival of the Vanities for DBAs | Pythian Group Blog — Fri, 22 May 2009 16:44:37 +0000

[...] Lewis has the perfect thing to follow your Log Buffer read—a lunchtime quiz on the Oracle Scratchpad. There are lots of answers already posted, so no peaking below the [...]

By: Row Directory « Oracle Scratchpad

Row Directory « Oracle Scratchpad — Thu, 21 May 2009 08:52:29 +0000

[...] under: Infrastructure, Troubleshooting — Jonathan Lewis @ 8:52 am UTC May 21,2009 Yesterday I asked the question: how many row entries can you create in an 8KB block with pctfree 0. It brought out some [...]

By: Jonathan Lewis

Jonathan Lewis — Thu, 21 May 2009 08:51:40 +0000

Adnras,
Good thought – but in extreme cases, Oracle is cautious, defensive, and wasteful. I may find time to blog about that some future point.

By: Adnras Gabor

Adnras Gabor — Thu, 21 May 2009 08:38:12 +0000

I have no DB instance to test this:

So nothing was said about specifying segment compression at the tablespace level.

I’d be interested if someone could test this.

Thanks,

By: Roberto

Roberto — Thu, 21 May 2009 08:04:26 +0000

I think Jonathan means, with “row entries”, the pointers into row directory. If I delete rows the row directory space doesn’t come back immediately. The next rows inserted don’t use (immediately) old pointers, hence “row entries” number grow.

By: Jonathan Lewis

Jonathan Lewis — Thu, 21 May 2009 07:41:01 +0000

Timur,

Interesting that there’s a little difference between freelist managed and ASSM – and there are other odd little details that show up for other reasons: try insert /*+ append */, try alter table t1 move.

The strangest thing with the ASSM one is that the amount of free space in the block shows that there is still room for one more row – did you find a way of making Oracle use that space ?

By: Jonathan Lewis

Jonathan Lewis — Thu, 21 May 2009 07:37:52 +0000

Martin,

The minimum space requirement for allowing migration to take place is important – and has an interesting side-effect – but it’s not the critical detail in this case.

By: Jonathan Lewis

Jonathan Lewis — Thu, 21 May 2009 07:35:25 +0000

Joel,

Definitely getting warm with that comment.

By: joel garry

joel garry — Wed, 20 May 2009 22:28:10 +0000

Since Oracle can reuse stuff in blocks, it ought to be unlimited. The question becomes, how efficient is oracle at reusing stuff in blocks. V-2 Schneider showed complete efficiency for indices.