Page 1 of 1
Problem with entire partitioning in lookup
Posted: Fri Dec 14, 2007 11:46 am
by RAJEEV KATTA
When I select entire partitioning in lookup and I have got 4 nodes with 4 matching rows I get the ouptut of 16 rows as it process all the data in each node.How to avoid this and get only 4 records.
Re: Problem with entire partitioning in lookup
Posted: Fri Dec 14, 2007 12:25 pm
by rjhcc
RAJEEV KATTA wrote:When I select entire partitioning in lookup and I have got 4 nodes with 4 matching rows I get the ouptut of 16 rows as it process all the data in each node.How to avoid this and get only 4 records.
select round robin.....
Posted: Fri Dec 14, 2007 12:28 pm
by RAJEEV KATTA
But the best partitioning is entire for lookup.
Posted: Sat Dec 15, 2007 2:32 am
by ray.wurlod
The best partitioning for lookup is identical partitioning to the stream input.
Posted: Mon Dec 17, 2007 7:37 am
by keshav0307
if the lookup stream is entire partitioned, then choose the main stream as auto, not as entire partition( i guess you have choose both input stram as enite and thats why your are getting 16 rows)
Posted: Mon Dec 17, 2007 8:09 am
by Minhajuddin
ray.wurlod wrote:The best partitioning for lookup is identical partitioning to the stream input. ...
Even when the input partitioning is round robin?
Partition both the input and the reference links using the Hash partitioning method on the key columns. This should fix your problem of getting duplicates in the output.
Posted: Mon Dec 17, 2007 1:52 pm
by ray.wurlod
I should have aid identical key partitioning. This implies either Hash or Modulus as the partitioning algorithm.