Does overfeeding reduce the query performance?
Posted: Sat Feb 19, 2011 10:59 pm
Hi,
I am trying to understand if overfeeding reduces the query performance in a cube that has rules.
Ideally my feeders should be conditional to avoid overfeeding, but to improve the time it takes for re-processing feeders (while loading cube into memory, or saving a rule) and to enable multi-threading for this, I used unconditional feeders which obviously resulted in some overfeeding. Since I do have a lot of memory in 64bit machine, I did not worry about additional memory requirement. But I could not achieve a very good performance for querying the cube, and I am wondering if overfeeding caused this performance degradation.
My hypothesis is that since I am feeding some cells that are not needed to be calculated (as any calculation will turnout to be zero for them), potentially I might have caused more cells than needed to be evaluated for calculation through overfeeding, which obviously reduces the query performance.
Any comments by experts will really be appreciated.
Regards,
I am trying to understand if overfeeding reduces the query performance in a cube that has rules.
Ideally my feeders should be conditional to avoid overfeeding, but to improve the time it takes for re-processing feeders (while loading cube into memory, or saving a rule) and to enable multi-threading for this, I used unconditional feeders which obviously resulted in some overfeeding. Since I do have a lot of memory in 64bit machine, I did not worry about additional memory requirement. But I could not achieve a very good performance for querying the cube, and I am wondering if overfeeding caused this performance degradation.
My hypothesis is that since I am feeding some cells that are not needed to be calculated (as any calculation will turnout to be zero for them), potentially I might have caused more cells than needed to be evaluated for calculation through overfeeding, which obviously reduces the query performance.
Any comments by experts will really be appreciated.
Regards,