Comments on: 11gR2

By: Dharmendra

Dharmendra — Mon, 14 Sep 2009 17:59:35 +0000

Does “Instance Caging” reserve CPUs as well as Memory (e.g. Process memory ) ?

Thanks

By: hpdba

hpdba — Sun, 13 Sep 2009 18:56:39 +0000

Jonathan,
I was interested when I saw Oracle’s marketing blurb:
“Oracle Database 11g Release 2 also offers the concept of ‘Instance Caging’ where databases can be confined to use specific cores in the SMP environment, thereby removing the need for virtualization or LPAR software.”
Looking instance cagine up in the manual reveals it just limits “the number of CPUs that a database instance can use simultaneously”, by foreground processes via resource manager and CPU_COUNT. That doesn’t sound the same as restricting the instance to specific cores.
I have never used RESOURCE_MANAGER_CPU_ALLOCATION, but perhaps 11gR1 did the same thing with this parameter, which is now deprecated.
Ari

By: Entradas de Oracle semanas 35-37 « Gruñidos sobre Oracle y SAP

Entradas de Oracle semanas 35-37 « Gruñidos sobre Oracle y SAP — Sun, 13 Sep 2009 11:41:48 +0000

[...] Jonathan Lewis: 11gR2 [...]

By: coskan

coskan — Fri, 11 Sep 2009 13:28:55 +0000

About the deferred segments Nicolas Gasparotto found a bug that it is not working with alter table move option

By: Top 10 Oracle 11gR2 New Features | Structured Data

Top 10 Oracle 11gR2 New Features | Structured Data — Wed, 09 Sep 2009 21:03:37 +0000

[...] catching up on blog posts I see that Jonathan Lewis, Christian Antognini and Nuno Souto picked up on the deferred segment creation new feature in [...]

By: Jonathan Lewis

Jonathan Lewis — Sun, 06 Sep 2009 10:43:45 +0000

Ari, "can't find any segments" - maybe the question "why isn't Oracle using my index?" will now be overtaken in the popularity race by "why isn't oracle creating my index?" Charles' posting isn't really tied very closely to 11gR2 - but a new release gives people a chance to do some proper testing, and there were some interesting obvservations that could be made about his results (which, at first sight, might have been attributable to some change in the Oracle code - until he pointed out his oversight with the filesystemio_options). I've changed the URL to pick up the whole thread, which includes the data generation code and the query. My comments were similar to yours, but I think there are still a couple of minor variations in performance to examine, even though the biggest variation was almost certainly related to the effects of a file-system cache (possibly coupled with a slightly unexpected sub-optimal use of the buffer cache).

By: hpdba

hpdba — Sun, 06 Sep 2009 06:20:20 +0000

Hi Jonathan,

I haven’t worked my way through the new feature list up to the one you’ve highlighted yet. Looks good, but I wonder if it will confuse someone when he/she creates a schema, without populating it, and wonders why no space is used and why he/she can’t find any segments :) Or perhaps the segments will be listed as having 0 blocks – I can’t test this yet.

BTW, I can’t see what Charles’ post has to do with 11gR2 specifically. The title is “Surprising Performance Changes with Oracle 11.2.0.1″ but the 11gR1 and 11gR2 results are very similar for the same settings.
I’d guess (from the avg I/O times) that the OS cache was pre-warmed for at least some of the tests, while the database’s cache was cold for them all (high physical I/O). That was a big factor behind why the direct I/O tests were so much slower than the buffered I/O tests.
I’d also say that the index was large and not appropriate for the SQL. (Full scan/range scan did three times more logical I/Os than the table size for 2.5% of the table’s rows). But then I don’t have the SQL, explain plan, etc etc.
What did you make of it?

Ari