Serialization and 2PL

From: Krishna <send2krishna_at_rediffmail.com>
Date: Fri, 4 Apr 2003 14:38:25 +0530
Message-ID: <b6jn0j$630f7$1_at_ID-187332.news.dfncis.de>



Hi ,

"locking" is an implementation technique for achieving Serialization.

Most of the books that I have come across (Gray's - TP concepts & techniques, Bernstein's - ConnControl & Recovery etc...) talk about how well-formed locked & two-phase locked transactions help achieve serialization. The proofs are also understandable. But they don't explain (or I didn't understand) how the concept of
"well-formed" and "two-phase" locked transactions was envisaged.
I mean the serialization theory is explained ... all that stuff about detecting cycles in dependency graphs etc ... but then soon after that books jump into explaining how locking solves the problem .... proving that tnxs having some special locking properties will form a serializable schedule. It's like they (authors) know something works and they explain how the principle works but theres no explanation of how they arrived at the technique/principle in the first place.

In other words, whats missing from the explanation is, given serialization theory how do we arrive at the locking principles which'll help one achieve serialization?

So could someone explain , or point me to the correct resource which gives this kind of description.

Regards,
krishna Received on Fri Apr 04 2003 - 11:08:25 CEST

Original text of this message