Row buffering option

vishnuvyas · Post by **vishnuvyas** » Mon Jan 05, 2009 12:24 pm

We have a option of enabling the row buffering in DataStage Administrator.

It gives us two options :
In Process and Interprocess and Admin guide suggests that it helps to improve the performance. I can't understand how this option increase the performance of DS jobs.

Can someone explain me the about this option in detail and when we should use it and how to optimize the cache size?

eostic · Post by **eostic** » Mon Jan 05, 2009 1:44 pm

This feature allows Server jobs to exploit degrees of parallelism. It has some caveats, and shouldn't be turned on "in general" but at times can be very beneficial. Do some searches --- I think you'll find a lot of detail on it in the site.

If you have any other detailed thoughts on it, start a thread in the Server forum...

Ernie

ray.wurlod · Post by **ray.wurlod** » Mon Jan 05, 2009 4:31 pm

In particular, setting of row buffering is entirely irrelevant in parallel jobs. Buffering is automatic, and tuned by a half-dozen or so environment variables with BUFFER in their names. Intra- or inter-process row buffering is only relevant to server jobs.

vishnuvyas · Post by **vishnuvyas** » Tue Jan 06, 2009 1:19 am

Thanks Ray. why do we have an option in paralllel jobs for setting these parameters ?

ray.wurlod · Post by **ray.wurlod** » Tue Jan 06, 2009 1:47 am

Read the on-line help to learn about this control. Essentially, though, it allows you to override the default behaviour. Be very careful in your reading to identify WHERE this buffering happens (it's probably not what you're thinking). Please post the results of your research here.