Prashant Narayani

• Nov 17th, 2006

 

in many ways we can improve,one simple method is  by inserting IPC stage between two active stage or two passive stages.....

sunilsavalasang

• Nov 22nd, 2006

 

there r lots of techniques for performance tuning as u asked ipc should b inserted btn two active stages ......................

amardeep.reddy

 Profile Answers by amardeep.reddy 

• Dec 6th, 2006

 

Some of the tips to be followed to improve the performance in DS parallel jobs-

1. Do right partitioning at right parts of the job, avoid re-partitioning of data as much as possible.

2. Sort the data before aggregating or removing duplicates.

3.  use transformer and pivot stages limitedly.

4. Try to develop small simple jobs, rather than huge complex ones.

5. Study and decide in which curcumstances a join or merge can be used and in which a lookup can be used.

prav2000

 Profile Answers by prav2000 

• Dec 21st, 2006

 

Instead of having a job with fork join create two jobs. These two jobs will perform better then the single job in most of the cases. Even if you want to you fork join style then use proper partitioning and sorting for the stages.

Bhaskar

• Dec 29th, 2006

 

We can put hash file in lookup. This will index the input data based on key column (which we define). Thus improve performance. Also Array size can be increased in final table stage.

veer.dwh

 Profile Answers by veer.dwh Questions by veer.dwh

• Sep 27th, 2009

 

There are several performance techniques are avilable to help run the job faster

1. Eliminate unwanted data and columns as early as possible in your job design, each unwanred column and data  requires extra cpu utilization and memory.
2. Eliminate unwanted datatype conversions which are not required.
3.Use tranformmer stage wisely, if you use more no of transformers it will decrease your job performance.
4.Use sorted input for agrregation to give good performance.