Archives of the TeradataForum

Message Posted: Fri, 08 Mar 2002 @ 14:07:00 GMT


	<Prev	Next>		<<First	<Prev	Next>	Last>>

Subj:		Re: Teradata locking mechanism vs. Oracle

From:		Sauli Harala

Yet another difference, - is the granularity of the (write) locking in Teradata - compared to Oracle.

Somewhere in the Teradata documentation there is quite profound locking summary table, describing the update select statement referred columns and the behviour if those referred columns are UPI, NUPI, USI, NUSI or no index.

As far as I understand this table correctly, you easily lock more rows with Teradata update statements than you would with Oracle row-level locking.

(This may be derived from the Teradata 'share-nothing' parallel architecture.)

Teradata is more focused to DWH than to OLTP.

To achieve what you've described below, you'd have to have App A store the changes in a separate table and then when it comes to the point of doing a 'commit' those changes would need to be applied to the real table.

I would consider this suggestion in case you make long term transactions with Teradata.

If you only have single update process and multiple read the Teradata 'dirty-reads' work OK - if they do not cause too confusing results to the reader processes ...


	<Prev	Next>		<<First	<Prev	Next>	Last>>