DataStage vs. Ab Initio

[ray.wurlod]

Don't take your eye off the "Cloud" ball. "They" are investing hugely.

[michaeld]

I disagree. Development time in Ab Initio is faster than in DataStage for any job that is moderate or more complex. Basic jobs may be faster in DataStage.

1) making reusable jobs in Ab Initio is easy. Reusable jobs in datastage are impossible if the job is is not basic. This makes development longer in DataStage for implementations that follow patterns.

Also, RCP does not work with all stages.

2) DataStage makes you come up with creative ways to implement functionaily with it's rigid stages, while in AbInitio most components can be extended in the gui by writing XFR code.

3) DataStage transformation language is too limited. In Abinitio you can write functions and package them for all to use. Building a framework of functions will speed up EDW development. In DataStage you have to write C++ code. Even then , deploying versions of reusable code to all projects is a maintenance issue.

[Abhijeet1980]

AI is good for companies having too many business entities and each one needing an ETL at its service.

The re-usability is certainly an advantage over DS for it differs very fundamentally in the following concepts:

1. Job design
2. Transformation and structure
3. Executable is a script

DataStage never differentiated this and mixed first two points into single job.

Disadvantages:

1. AI is a huge investment and not much expertise is available.
2. Not much documentation is available with people or on internet.
3. AI was initially developed for military purpose applications and not much licensing is available or made available to small companies.
4. AI is substandard (My opinion).
5. AI develops stages that customized for client companies and hence followed no standard till date. (Supports my prejudice)
6. Hence more error prone. (Absolute inference)
7. Performance wise - DS and AI are more or less equal. (No bias)
8. Ease of use - Go for PowerCenter.
9. Highly sophisticated if you go for EME alongwith which is kinda version control along with a file system based repository. With 2.14/1.14, there was web based console meant for monitoring the jobs or scripts.

Hope, that was enough else a huge list of "Why avoid AI" is there to be published soon.

Many regards

[rameshrr3]

This is such an old thread , and has some useful , though possibly dated info .

Given what i know about Ab-Initio - It looks like the Apple Macintosh of ETL.

Most companies do not want to invest in Ab-Initio because developers are scarce and costly- I guess they (Ab-Initio) may have some learning to do .

Does this article hint anything about the wall of secrecy around ab-initio ?

http://www.inc.com/magazine/19950915/2622_pagen_4.html

[nagarjuna]

In my opinion ,

Abinitio is more powerful if you are having very complex rules . As mentioned by many posters , Datastage is easy to learn & more developers are available in the market.

Coming to administration point of view , Datastage has many hidden or esoteric things going in background . The architecture is a mix of both the file system & database . There are many internal tables in datastage repository which are little difficult to understand .
Runtime binding ( Parameterizing the transformations ) is possible in Abinitio but not datastage .

Finally , If you are good at shell scripting then you would appreciate Abinitio compared to datastage .

[vmcburney]

I like Ab Iinitio, I like the way they play their own game on marketing and sales and don't mimic everyone else. For a few years there they were the fastest ETL tooling by a big margin until Ascential Software picked up Orchestrate and switched to a parallel engine. Today the gap is smaller and Ab I may be faster - but the cost means I rarely come across it and there are very few developers available. The one customer I have come across who bought it were surprised at the phenomenally high cost of building just two transformations - flying developers into Australia from the US isn't cheap.

For very high data volumes I think the approach should be a combination of DataStage, CDC and Streams. I think Streams may be more similar to Ab I than DataStage with an emphasis on custom built transformations.

[nagarjuna]

Vincent ,

Just curious what is a stream ? Is it another ETL tool ?

[ray.wurlod]

Vincent is referring to InfoSphere Streams, another IBM product designed to process data "on the move". Research this at IBM website.