Stateless clients and locking schemes (or rather isolation levels)

From: Serge Rielau <srielau_at_ca.ibm.com>
Date: Wed, 27 Jul 2005 07:17:00 -0400
Message-ID: <3kp8pgFvghd4U1@individual.net>

Superboer wrote:

>>This is a funny way of looking at. Obviously Oracle's none locking
>>engine is perfectly suited to scaling multi user applications,
>>particularly when most people are developing for stateless clients.

>
>
> ahum does the above explain why informix was faster on a 5 times
> smaller machine then obstacle...????
>
> Superboer.
>

Changed the subject lines and following up on what Knut started.

How does Oracles snapshot isolation help with stateless clients. To the best of my knowledge snapshot semantics only operate on either a statement or a transaction level. In a stateless scenario I'd assume that teh application transaction covers at least two database transactions. A read phase wher the resultset is displayed at the client and a separate write phase where the modified data is written back. How does snapshot isolation help here?
Informix supports versioning columns which can be used by the app to prevent overwriting other users changes across DB transaction boundaries. MS SQL server has a somewhat similar approach and even buried optimistic locking into the cursor logic (not applicable in a stateless enviroment (no cursor open).
I see tha value of snapshot isolation for certain purposes. I don't see it for a 3 tier web application....

Thoughts?
Cheers
Serge

-- 
Serge Rielau
DB2 SQL Compiler Development
IBM Toronto Lab

Received on Wed Jul 27 2005 - 06:17:00 CDT