Comments on: Consistent Gets – 2

By: 一致性读2-翻译jonathanlewis的一篇blog | dbafree首页

一致性读2-翻译jonathanlewis的一篇blog | dbafree首页 — Tue, 28 Jun 2011 02:06:44 +0000

[...] 翻译自：http://jonathanlewis.wordpress.com/2009/06/12/consistent-gets-2/ [...]

By: What is consistent gets and db block gets? | Xing AiMing's Home Page - 邢爱明的个人网站

Wed, 15 Jun 2011 07:00:53 +0000

[...] References: Quick Tips Consistent Gets – 2 [...]

By: Oracle: Checking Object Index Leaf Blocks «

Oracle: Checking Object Index Leaf Blocks « — Sat, 13 Nov 2010 05:44:34 +0000

[...] have better execution plan against the Cost-Based. Using RULE vs CHOOSE the result of the consistent gets is 70% smaller than the CHOOSE. [...]

By: Scalability Conflict « Oracle Scratchpad

Scalability Conflict « Oracle Scratchpad — Fri, 30 Jul 2010 17:52:21 +0000

[...] of index blocks. The process running the query will do a lot of work reading undo blocks to create read consistent copies of the blocks it needs to [...]

By: True or False – Undo « Charles Hooper's Oracle Notes

True or False – Undo « Charles Hooper's Oracle Notes — Sun, 09 May 2010 14:10:48 +0000

[...] http://jonathanlewis.wordpress.com/2009/06/12/consistent-gets-2/ [...]

By: Statspack/AWR Report Resources « Charles Hooper's Oracle Notes

Statspack/AWR Report Resources « Charles Hooper's Oracle Notes — Mon, 14 Dec 2009 14:14:04 +0000

[...] #4 An absolutely confusing definition of consistent gets, which are different from current mode gets. The consistent get, while coming from memory, may have involved one or more physical disk reads. The “db block gets” statistic is actually the same thing as a current mode get. A consistent get attempts to build a version of a block as of a specific time, or more accurately, as of a specific SCN through the application of zero, one, two, or more undo blocks. The specific SCN to which the block is rolled back is dependent on the current isolation level specified for the session. A current mode get will contain both committed and uncommitted data since it is the version of the block as it exists at that instant. See these blog posts: http://jonathanlewis.wordpress.com/2009/06/09/quiz-night/ http://jonathanlewis.wordpress.com/2009/06/12/consistent-gets-2/ [...]

By: Nigel

Nigel — Mon, 13 Jul 2009 13:43:19 +0000

That is very intersting! It also blows apart my understanding of how Oracle works! I had, rather simplistically, believed that SCN was King and that first past the post was the “winner” in transactions. I had no idea that (relatively) complex stuff like that you describe went on. Is this in the Oracle Docs.?

By: Jonathan Lewis

Jonathan Lewis — Mon, 29 Jun 2009 18:34:10 +0000

Nigel,

I’ve modified the wording to clear this up. Your session is allowed to run code behind the scenes that sees the uncommitted changes even though you, as an end-user, don’t have access to those changes.

By: Nigel

Nigel — Mon, 29 Jun 2009 15:55:48 +0000

“you need to see uncommitted changes from other sessions in case you are trying to make a change that will become illegal when the other session commits.”

I am confused! I thought that uncommitted changes in other sessions were absolutely not ever visible. I am missing something glaring, I know, but what?!

By: Glossary « Oracle Scratchpad

Glossary « Oracle Scratchpad — Sat, 27 Jun 2009 20:12:59 +0000

[...] LIO: (a.k.a Logical I/O) – a visit to a block that is in the buffer cache, initiated by a search though a cache buffers chain while holding a latch. (See also PIO). For further reading about logical I/Os, see this note. [...]