Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Mailing Lists -> Oracle-L -> Asynchronous Commit Feature of 10.2 - Basic Qs

Asynchronous Commit Feature of 10.2 - Basic Qs

From: VIVEK_SHARMA <VIVEK_SHARMA_at_infosys.com>
Date: Fri, 22 Sep 2006 13:40:42 +0530
Message-ID: <BBD944BCAC3AB4499DFBAFB1D8AF302001D3E27D@BLRKECMSG11.ad.infosys.com>

Folks

Qs 1 When using COMMIT_WRITE=BATCH,NOWAIT, After a database crash, if the respective Application has the capability to check within the database for

volume of transactions lost after complete recovery & re-play the same, Can the COMMIT_WRITE=BATCH,NOWAIT feature be used in such a case?

This basically refers to periodic Batch nature of Transactions which run with NO intermediate User intervention.

Qs 2 For a Process/SESSION consider that COMMIT_WRITE is set to BATCH,NOWAIT. If 2 Commits are issued sequentially (i.e. one after another), is there a possibility that the Commit which was issued later may be written 1st to the

hard Disk ? Personally I think the answer is "NO"

Qs 3 If multiple processes are running parallely & committing data with a related logical inter-dependency between them while doing updates, can the

same be maintained when using the Asynchronous commit feature? To Explain this in simpler words:-

Assume for 1 process COMMIT_WRITE is set to BATCH,NOWAIT.

Assume at 10:00 this process issues a Commit after updating a field value to

say "A".

Assume that Oracle decides to write it to redo logs on Disk at 10:02.

Question - If at 10:01 ANOTHER process reads(SELECTs) the same field value,

will it get the UPDATED field Value of "A"

Qs 4 Is there any difference in the crash recovery mechanism in general with the use of this feature when compared to not using the same, both for Single instance & RAC setup?

Qs 5 Is there any difference in the rollback mechanism in general with the use of this feature when compared to not using the same both for single instance & RAC?

Qs 6 For COMMIT_WRITE=BATCH what are the general guidelines for data to be batched & written to disk? Is there an outer-time line involved? Do writes to disk necessarily happen at some periodicity?

Qs 7 If archivelog mode is enabled, is there any change in the WRITEs mechanism to the archived redo log files?

Qs 8 Performance benefits seen when using this feature?

Qs 9 Any Good Links / Docs on this?

Thanks indeed

Received on Fri Sep 22 2006 - 03:10:42 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US