I am working with a client, building a cube with more than 100 million records. We are trying to improve build time any way we can and have implemented most of the standard Cognos suggestions.
The database is on a SQL server and indexing the transactional file has been suggested as a method to increase build time performance. I'm skeptical that this will make much, if any difference.
Does anyone have suggestions as to whether indexing the transactional files is worth pursuing?
Indexing the transaction file will result in Cognos reading the data faster so it is definately a good suggestion. It should reduce the build time of the cube.
Some enhancements have been made to the in-memory processing of Transformer during cube builds for the Cognos 10 release. Is it an option for your client to upgrade?
MF.
Once we add the index, I assume I need to reference the same index in the data source. Is this correct
We are in the process of moving to Cognos 10, but it probably won't be implemented until after the first of the year. We are rebuilding the data warehouse at the moment and trying to squeeze every bit of optimization we can out of both the data warehouse and model designs now.
MFGF, question for you...if you were using a package as data source in Transformer then to leverage the use of table indexes, you must check mark that option when importing query subjects on FM, right? and doing so will instruct Cognos to use the index for columns in that query subject i.e. table.
my question is, does this also port over to Transformer when using the package, i.e. does Transformer recognize the same way FM would that it's supposed to use the indexes for the columns that have it?