IBM Cognos BI Tuning

The way that you use and configure IBM Cognos BI can influence its performance. For instance, you can design models and reports with performance in mind, configure IBM Cognos BI dispatchers and services for performance, and schedule jobs to make the best use of system resources.

Designing Models and Reports for Performance

Designing and creating models in Framework Manager is an essential step in the IBM Cognos BI workflow. A model specifies, structures, adds to, and manages the metadata used to create reports. For best IBM Cognos BI performance, a modeler can design models that specify default prompting, and that set the query processing type.

Specifying Default Prompting

To improve query performance, models can include reports that prompt a user for a descriptive name and perform filtering on a code or key value. We can modify a model to make sure that queries sent to the data source are efficient, well formed, and secure. To ensure optimal performance, the Framework Manager modeler can

• specify the rules governing query generation  
• restrict user access to specific rows or columns of data
• model data relationships to hide the complexity of data from report authors

Setting Query Processing Type

For relational metadata, We can develop performance by selecting the right type of query processing. There are two types of query processing:

• limited local

The database server does as much of the SQL processing and execution as possible. However, some reports or report sections use local SQL processing

• database only

The database server does all the SQL processing and execution. An error emerge if any reports or report sections require local SQL processing.

Although the database server can usually run the SQL and run reports much faster, local processing is sometimes necessary. For example, choose limited local processing if you want to create cross database joins or if you want report authors to use unsupported SQL99 functions. Some complex queries, such as a query that must generate an At clause to avoid double-counting, require limited local processing. In this case, the query automatically uses limited local processing even if the package was published with database only processing.