What does it get better performance? Filter or transformer?

Post questions here relative to DataStage Enterprise/PX Edition for such areas as Parallel job design, Parallel datasets, BuildOps, Wrappers, etc.

Moderators: chulett, rschirm, roy

Post Reply
manuel.gomez
Premium Member
Premium Member
Posts: 291
Joined: Wed Sep 26, 2007 11:23 am
Location: Madrid, Spain

What does it get better performance? Filter or transformer?

Post by manuel.gomez »

I have found this reading some Ascential documentation:
Transformer constraints are FASTER than Filter stage!
Filter constraints are interpreted
Is it true? I have always tryed to avoid transformers and using filter stages instead, but now it seems that transformer is better!
ArndW
Participant
Posts: 16318
Joined: Tue Nov 16, 2004 9:08 am
Location: Germany
Contact:

Post by ArndW »

I would test this empirically before taking the documentation's word for it.
mdan
Charter Member
Charter Member
Posts: 46
Joined: Mon Apr 28, 2003 4:21 am
Location: Brussels
Contact:

Re: What does it get better performance? Filter or transform

Post by mdan »

manuel.gomez wrote:I have found this reading some Ascential documentation:
Transformer constraints are FASTER than Filter stage!
Filter constraints are interpreted
Is it true? I have always tryed to avoid transformers and using filter stages instead, but now it seems that transformer is better!
Yes, at least for 7.5.2 transformer is faster. You may use transformer unless the special situation when the column that you filter by should be defined (parameter) at the runtime.
John Smith
Charter Member
Charter Member
Posts: 193
Joined: Tue Sep 05, 2006 8:01 pm
Location: Australia

Re: What does it get better performance? Filter or transform

Post by John Smith »

manuel.gomez wrote:I have found this reading some Ascential documentation:
Transformer constraints are FASTER than Filter stage!
Filter constraints are interpreted
Is it true? I have always tryed to avoid transformers and using filter stages instead, but now it seems that transformer is better!
Don't be overly concerned which one has better performance, I think it's safe to use both and if you job is written properly both stages will perform satisfactorily.
sreddy
Participant
Posts: 144
Joined: Sun Oct 21, 2007 9:13 am

Re: What does it get better performance? Filter or transform

Post by sreddy »

Hi

That is quite natural, we can do
1. Filter the data (Constraint)
2. Do conversions
3. We can create Surrogate Key
So many things we have to do. That is why performance is low.

Filter stage is specified only for that purpose

http://it.toolbox.com/blogs/infosphere/ ... -evil-9801
manuel.gomez wrote:I have found this reading some Ascential documentation:
Transformer constraints are FASTER than Filter stage!
Filter constraints are interpreted
Is it true? I have always tryed to avoid transformers and using filter stages instead, but now it seems that transformer is better!
SReddy
dwpractices@gmail.com
Analyzing Performance
chulett
Charter Member
Charter Member
Posts: 43085
Joined: Tue Nov 12, 2002 4:34 pm
Location: Denver, CO

Re: What does it get better performance? Filter or transform

Post by chulett »

Ah yes, a Vincent blog. :wink:

Keep in mind the fact that his observations are just over two years old now and - from what I understand - the stage has since been vindicated, rewritten to dramatically improve its speed.
-craig

"You can never have too many knives" -- Logan Nine Fingers
vmcburney
Participant
Posts: 3593
Joined: Thu Jan 23, 2003 5:25 pm
Location: Australia, Melbourne
Contact:

Post by vmcburney »

I would avoid the filter stage - it's not as user friendly as a Transformer and you are likely to be using your Transformer for multiple things. The one time I found a filter stage shaved about 5% off my job run time was when I was filtering out over 75% of the incoming rows and feeding the data to a Transformer with a lot of transformation logic. The Filter stage avoided a lot of unneccessary transformations - but then maybe two Transformers in a row would have been even faster.
John Smith
Charter Member
Charter Member
Posts: 193
Joined: Tue Sep 05, 2006 8:01 pm
Location: Australia

Re: What does it get better performance? Filter or transform

Post by John Smith »

sreddy wrote:Hi

That is quite natural, we can do
1. Filter the data (Constraint)
2. Do conversions
3. We can create Surrogate Key
So many things we have to do. That is why performance is low.

Filter stage is specified only for that purpose

http://it.toolbox.com/blogs/infosphere/ ... -evil-9801
manuel.gomez wrote:I have found this reading some Ascential documentation:
Transformer constraints are FASTER than Filter stage!
Filter constraints are interpreted
Is it true? I have always tryed to avoid transformers and using filter stages instead, but now it seems that transformer is better!
The reference to that blog is dated 2006 and we're halfway through 2008. come on that information is already dated and you should have some faith that IBM has done some work with the product during all that time. Don't think they pay their developers to sit idle... :roll:
Post Reply