Benifit of using Parallel transformer

[arsheshadri]

Hi,

Earlier we had some issues with C++ compilers and could not use Parallel transformers in jobs.

But now the issue is solved and we have an option to use the parallel transformers.

In all our existing jobs, we have been using BASIC transformers.
Now because of availability of parallel transformer we are thinking of replacing the BASIC transformer with parallel transformer in all our existing jobs.

I have some doubts regarding the above issues -

1. Is it a right idea to replace BASIC transformer to parallel transformer?
2. Are there performance impact / gains if we do so?
3. Are we missing any major features by using BASIC transformer?
4. Is there any particular things that is to be taken care while replacing the transformer?
5.One of the problem with our job is, we have used ICONV and OCONV in current transformer, which I think will not work in parallel transformer and we need to replace some coding also. So is it worth replacing?

Thanks for your time and suggestion.
Sheshadri

[ashwin141]

Hi

Here are the possible answers to your questions.

1. Is it a right idea to replace BASIC transformer to parallel transformer?
Yes if you are working in PX. It will improve your performance as the parallel transformer will process the data parallely and there are many functions available in parallel transformer which are not there in basic transformer.

2. Are there performance impact / gains if we do so?
Using parallel transformer may affect your job performance negatively, its recommended only when you can't do the same job with any other stage in PX. Incase u have to implement some complex business rule its recommended.

3. Are we missing any major features by using BASIC transformer?
Not much.

4. Is there any particular things that is to be taken care while replacing the transformer?
Depends on your job design and requirements.

5.One of the problem with our job is, we have used ICONV and OCONV in current transformer, which I think will not work in parallel transformer and we need to replace some coding also. So is it worth replacing?

Again it depends on your requirements.

That's all I can say. Others may throw some more light on it.

Regards
Ashwin

[ray.wurlod]

I would have to add to the answer to point 2 that the parallel Transformer stage will always out-perform the BASIC Transformer stage achieving the same task, for a reasonable volume of data.

This is because to use the BASIC Transformer stage it is necessary to convert the stream of data into the "typeless" environment of DataStage BASIC and back again.

There is no direct equivalent to Iconv() and Oconv() in the parallel Transformer stage, but there are many conversion and casting functions available that ought to be able to do most of the common tasks (such as date reformatting) encountered with Iconv() and Oconv() in server jobs. Because there are data types (such as Date) in parallel jobs some of these conversions may, indeed, be unnecessary.