how can i improve performance in Aggregator

chanthou_2000 · Post by **chanthou_2000** » Thu Jul 21, 2005 4:37 am

hello,

i'm looking for some ways to improve performance of Aggregator Stage.
But i didn't find any options to say to DS that my source is already sorted (i use an order by sql clause) like in DS 5.0 Server where there was such options (i used to use it !)

Thank you

Chanthou

thebird · Post by **thebird** » Thu Jul 21, 2005 5:40 am

Chanthou,

There is an option in the sort stage- sort mode, where you can sprecify : "Don't Sort (Previously Sorted)".

Hopefully this will take care of your issue.

Regards,
The Bird.

ray.wurlod · Post by **ray.wurlod** » Thu Jul 21, 2005 4:45 pm

The options for specifying sorted input are in the Columns grid in server job Aggregator stages.

In parallel stages, sorting of input is managed as a property of the input link; I don't have access to PX at the moment to give you specifics, but it's certainly possible.

s1kaasam · Post by **s1kaasam** » Thu Jul 21, 2005 9:10 pm

In the aggregator stage there is in the option section wherein you could select the method to be sort instead of hash.

Now before you run into this aggregator use a sort stage and sort it on the the grouping keys of the aggregator.

The partitioning into the sort stage hash it on the grouping keys and into the aggregator keep it auto.

try this and I am sure you would have better performance gain.

Any questions let me know.

kumar_s · Post by **kumar_s** » Fri Jul 22, 2005 2:14 am

Is it really must to do a sort using a sort stage before an aggregator everytime, bcoz when i tried with an un presorted data it worked same.
In the other case, why cant we maintain the partition as "Same" in Aggregator followed by sort stage.