Well, 20 mln of data doesn’t seem big enough to start causing troubles on Ms SQL 2005. Although your “circular joins” are a bit worriyng — there shouldn’t be anything like circular joins in well-designed reporting scheme. If it’s possible, I’d like to look at your current design, to see if there are are any improvements at database level..

10 mins of data processing for query isn’t reasonable time (although it depends on your hardware as well) to do analysis, 5s-1m is only interval possible.

DMRs are pretty good on performance side, when you’ve got a good DBA nearby to tune up the SQL server itself and your FrameWork manager scheme is pretty straight-forward (just a simple star, for example). In other cases, it may turn out usefull to build some aggregate (in your case, full) cognos powerplay cubes to speed up query processing.

- Yuri.

We are trying to build out a DMR with Cognos 8.3 and MS SQL 2005 (32 bit). Our consultant suggested the idea of a DMR and said that performance would not be a problem. Our consultant then quite to work for another firm and we are left in the lurch to figure this out on our own.

Our fact table is currently about 20 mln records; 8 years of data. If you joined all the tables as they are used by the OLTP, you end up with a pretty big snowflake with several circular joins.

We have imported that tables from a replicated copy of the OLTP, created views on the database and have imported the tables and views into FM. The views were built anywhere we created a snowflake in our model so that our model is a star schema in FM.

We have not dimensionalized the data yet beyond this - no hierarchies.

When we run a query agains a years worth of records (about 2 million records), it can take up to 10 minutes to return.

Is this normal, or are we missing something?

Any ideas on how to improve the performance?

Thanks,

Bob

Seen you on Communities as well.

I’ll add your suggestion to this post

Thanks for reading )

I work in Finance and one of the uses we have for event studio is automated weekly, monthly reporting.

New Hire Requisition reports, Actuals vs Budget for Operating Expenses and Capital Expenditures etc. Purchase Req’s over a certain amount are setup in a report and automated so that they can be reviewed on a timely basis with upper management.

Thanks for posting this wealth of information!

In my experience, it only took some minor tuning (switching to native sql for fact tables) and then Star transformations worked brilliantly. I’m currently using 10mln row fact tables with star queries and it works pretty nice.

Can you share some details on your configuration (oracle + cognos versions) and generated sql?

Yuri.

Have you had any issues with DMR performance?

The Cognos SQL rewrite in Analysis Studio means that the query never utilises star transformation, and therefore causes ‘nested loops’, which have a knock on effect on perfomance.

Regards
Rob

You can use process explorer to see trnsfrmr.exe properties, where one of the file handles will be the model used.