Tips to improve Impromptu report performance? 3

[bianalyst]

Hello,

Do you have any tips on improved report performance to keep in mind during the design stage of Cognos implementation? Is there an option within Impromptu to do server-side processing to improve performance? Can you explain what is actually happening when such feature is used? Am I compromising other functionalities if I were to turn on this feature?

Any other tips in optimizing report performance would be greatly appreciated. Wanted to be proactive with the design rather than putting out fires come production time.

Thank you kindly!

 

[Shek]

Click report--> Query--> Client Server-->Database Only.
Yes you lose the Impromptu functions use which you can do only in flxible or limited local processing

When I grow up, I want to be a child

 

[sln007]

Our experience is: Use "Flexible Processing".

 

[hikabir]

Use "Database only" option for only server side processing. However, chances are you will need some Impromptu functions in your reports, so the next best is "Limited local processing", this pushing as much processing as possible to the server. Additionally you can look up the pdf under Help->Books Online->Mastering Impromptu reports->Chapter 6 Optimizing Performance for tips such as not putting summaries in headers, . . .

 

[BruceReed]

Hikabir's suggestion to read the chapter 6 is very good. There are many things to consider when optimizing, and it depends on your exact setup & requirements.

Flexible processing is USUALLY the best setting, as Impromptu will all the processing that it can at the server, and then pass it the PC.

However YOU have to be careful on setting up the report, as you might be doing thing inefficiently. Things like sorting & grouping Impromptu, when all you want the report for is loading into transformer/ PPES, will slow you down.

Hope this helps
Bruce

 

[benbotektips]

I found that the best thing to do with Impromptu, is let it do as little thinking (read:calculations) as possible, and work it like a data echoer - good for formatting and presentation, not deep thinking. In other words, take care of the data at the database level. This could mean creating a view that outer joins 2 tables, and using the view fields in Impromptu, instead of doing the outer join in the Impromptu catalog. Also, no report picklist, strictly catalog & file (u can do needed calculations in a catalog field for the picklist prompt, or even better the data table or view, again). Be specific in the data views, for example if a report's audience is going to be eastern U.S. district, throw in a view that will only include Easterns, for example, to give Impromptu a smaller dataset to work with. If u must use a function, make it a database one instead of Impromptu one. Finally, although a star schema database is great, if u create a report that spans >6 tables, it's almost guaranteed to spin, & u end up staring at a blank screen for a while. It sometimes helps to split a big report into 2 smaller ones.

 

[gset]

Hi all,

My tip concerns cross tables: never use them!

e.g. if you have a table with columns:
1) Year
2) Country
3) Sales USD

And you want Country in lines and Year in columns just do the following for column "Sales 2003":

Total(if (Year = 2003) Then (Sales USD) else (0))

Repeat this operation each time you need it.

IT IS REALLY MUCH MORE PERFORMANT, MORE MAINTENABLE AND VERY, VERY POWERFULL.

Hope this help.



Graham.

 

[b1ffa]

definately use flexible processing, but also examine the sql. if you have developed reports by re-using old ones, you can often remove fields that are not absolutely necessary. In otherwords try to keep the number of tables referenced to a minimum.