Architectural Question: 1

[CognosTalk]

What’s the average size of query and results that are returned back when a user requests some data from Cognos. I know it is hard to size when you have different kinds of data.

We have an Oracle database and the biggest table has like two 2.7 millions of record. it is estimated to have approximatelly 3 million records in tables in the future.

 

[blom0344]

For facts, the dataset returned will be a aggregated set (i.e. aggregation has already been performed at the database)
Unless you have a zillion of additional attributes from dimensional tables, the resultset should be quite a lot smaller than 2.7 million

Ties Blom

 

[CognosTalk]

Thank you so very much for your answer!
What do you think is better?
1. Extract sql from cognos and run it against the database to see how many rows it returned
2. Go to session manager and extract the query produced by database and see how fast it executed and how many rows it produced.
What do you think should be my approach in terms of estimating?
Maybe it is better to run the database query (that is produced from the cognos query) and estimate from there

 

[blom0344]

Absolute top is to run a trace at the database and capture the SQL that is actually executed (in some cases the database manager will do a rewrite when it thinks it can do better )
This will also give you a clue about the real execution plan. Executing this SQL from a DBA tool will net the result set (actually the explain plan should probably give the correct number of rows that will be returned)

Ties Blom

 

[CognosTalk]

It worked out just fine,
I was able to gather all hte executed information from oracle session logs and from Toad (# of rows)
Thanks for all your help!