IBM Cognos Business Intelligence 10.2.2 brings “distributed bursting” which allows a bursted report to be executed across multiple servers. This welcomed feature provides marked performance improvements due to the parallel execution now leveraging all available resources.
So How Does Distributed Bursting Really Work?
When a distributed burst report is run, a new query is created based on the report burst group query. For instance, if the report is burst by Product Line, burst keys might be Camping Equipment, Golf Equipment, Personal Accessories, Outdoor Protection, Mountaineering Equipment, and so on.
Report execution requests are created each for a range of burst keys. For example one request might be for Camping Equipment and Golf Equipment, the next for Outdoor Protection and Mountaineering Equipment, and so on. The number of keys assigned to each request is determined based on a number of factors such as the number of servers, the number of threads, and so on. These report execution requests are distributed to the available servers for parallel execution. Each request runs effectively as a separate, non-distributed burst report rendering a portion of the original burst. Outputs from all requests are stored with the original burst report.
How Do You Control Distributed Bursting?
Distributed bursting is enabled by the new report bursting property Run in Parallel which has three settings:
- Default – use the system default setting
- Enabled – override the system default setting to use distributed bursting
- Disabled– override the system default setting to not use distributed bursting
Distributed bursting was enabled by default in Cognos 10.1.1 but this default setting has now been set to disabled in Cognos 10.2.2.
How Do We Determine When to Use Distributed Bursting?
As distributed bursting runs portions of the report execution in parallel, this means that there will be more queries executed against the underlying query database. It must be ensured that the query database is capable of supporting the increased load.
In order to run the burst reports in parallel on multiple servers, the burst keys in the parallel bursting process are built-into SQL as literals in the filter.
Distributed bursting, by design, leverages to the max the resources available; a single burst report could consume all available execution threads on all servers thus denying execution threads to other reports. For that reason, one should consider running distributed burst reports in a lower priority to allow other reports to execute normally; priority is set when scheduling the report execution.
In terms of performance gains with distributed bursting, your mileage may vary. There are too many variables to predictively make comparisons between distributed and non-distributed report bursting. It depends upon your servers’ resources, database type, indexing strategy, report design, and so on.
With careful planning and preparation, the distributed bursting feature of IBM Cognos BI 10.2.2 can often be used to take advantage of burst report performance and to maximize throughput for mass report deployment.
Read more about the new and improved features of IBM Cognos Business Intelligence 10.2.2 offers here.
Does your organization make use of or has used distributed bursting? Share your experiences with this feature with us.