Using the standard kernel/fact concept we have found the performance of the prompt engine to be lacking. All our research shows that in a derived universe containing the facts and a kernel containing the dimensions and conditions for those dimensions the engine reads the kernel in at least once for every prompt in the report.
It is also clear that the size of the kernel universe is a major factor as well, in testing we have removed the descriptions unless absolutely needed and made other modifications to the kernel universe. In making all these changes we have been able to achieve a 690k kernel rather than a 1.5mb kernel.
We then ran a series of tests and found that performance was degraded each time a report had a condition added. Typical measurements were shown to be in the range of the following:
Original 1.5mb kernel 130 seconds
Reduced 690k kernel 70 seconds
Included/no kernel 45 seconds
These timings are rough and are ONLY the time it takes to load the prompt engine. Note that the timings are different in full client vs WEBI but the differences are relative to the same savings. This leads us to believe the issue is throughout Business Objects.
Has anyone out there had similar results?
Have you found a solution?
Thanks.
Greg
cablegp (BOB member since 2003-06-04)