Best partitioning method

[jpraveen]

Hi All,

my requirement is ,i am using join stage and also look-up stage s in my jobs,what is the best partitioning method for Look-up stage as well as Join stage . i am running on 2-node config file.
by default i am using auto-partition
though we give this as Auto -partitioning,what type of partitioning will take the datastage internally?

[samyamkrishna]

Use any key based partitioning method.

[jpraveen]

HI

what kind of key based partitioning method we should use,
i tried with Entire Partitioning,but the records are doubled,and also i suspect we cannot use round-robin,because some records will go into 1st node and some will go into 2nd node,
can anyone explain on this ?

[samyamkrishna]

use hash partitioning and use the the keys which you are using in join for the partitioning.

[Ravi.K]

Entire is non Key partition method.

If you have single key column and that is Integer then use Modulus otherwise Hash Partition.

Find the Key and Non Key Partition methods.

Key:
-----
Hash
Modulus
Range
DB2

Non Key:
---------
Round Robin
Random
Range
Entire.

[ray.wurlod]

That's too simplistic. You not only have to use a key-based algorithm, but you must also make sure that the partitioning is based on (at least the first) join key and that the algorithm will yield sufficient distinct values to spread data across all the available nodes.