Page 1 of 1
StatsByCube
Posted: Wed Sep 23, 2009 7:50 pm
by jorelb
I am trying to optimize a cube that pulls data from other cubes using only rules. Using performance monitor, I get the output below for }StatsByCube.
Number of Stored Calculated Cells 5,290,316
Number of Fed Cells 46,628,304
Memory Used for Calculations 64,283,392
Memory Used for Input Data 153,220,096
Total Memory Used 217,503,488
Can someone help me interpret the output? Also, what bothers me is that it has153,220,096 of Memory Used for Input Data but there are no entries in this cube.
Any help would be greatly appriciated.
Re: StatsByCube
Posted: Thu Sep 24, 2009 7:36 am
by jim wood
I'm only guessing here but I would say that the input data is the data coming from other cubes. This would tend to tell me that either the cube is dominated but rule geenerated values or your calculation might need to be optimised. (Feeder memory usage as far as I am aware would show against the source cube. you may want to check these.) Have you considered replacing some of the rules with a data transfer via TI? This effectively pre-calcs some fo the values.
Re: StatsByCube
Posted: Thu Sep 24, 2009 7:06 pm
by jorelb
I agree, I would prefer TI, but the main user requirement is real time calculations (they do not want to run the risk of forgeting to run a process). However, I did modify some rules and fixed some intra cube feeders from 1 source cube and the changes improved the intial view load time from 45 sec to 20 sec. The stats from the source cube was the same before and after the feeder change but the stats on the subject cube improved significantly. Below are the new stats of the subject cube.
Number of Stored Calculated Cells 294,126
Number of Fed Cells 27,331,505
Memory Used for Calculations 3,465,984
Memory Used for Input Data 55,243,776
Total Memory Used 58,709,760
What is interesting is that I was expecting only the "Number of Fed Cells" to change but all other measures significantly changed. I validated the system output and they were the same before and after the change.
Any thoughts on why the other measures changed?
Thanks.
Re: StatsByCube
Posted: Fri Sep 25, 2009 7:23 am
by jim wood
I can only gues that you were initially over feeding. It might be worth checking some others unless you are happy with perfermance now?