Hi All,
in our cube we process the dimensions on the following way: processupdatefor the specific dimension and processindex for the measure groups.
This would be good, but one of the dimensions make some problems (it is a hierarchical dimension, something like a tree with its leaves, etc.). If an old value of the dimension has been modified and if this value also effects many of the records of the fact table, than the entire indexes/aggregations of the measures get dropped and the processindex takes lot of time (that is not the case with any other "normal" dimensions).
I have found some thresholds in the config file that probably could block the drop of all of the indexes/aggregations, though I am afraid to modify them (they are still on the default setting).
AggIndexBuildThreshold and IndexBuildThreshold
Do you have any experience with them? Would you, guys, suggest to set these thresholds on a higher value or to null not to drop the indexes/aggregations? If they wouldnt be dropped would be the cube still giving back correct numbers after some old value in the dimension has been changed?
Thanks in advance for your answer!
additional info: sql server 2008 standard x64, win server 2008
in our cube we process the dimensions on the following way: processupdatefor the specific dimension and processindex for the measure groups.
This would be good, but one of the dimensions make some problems (it is a hierarchical dimension, something like a tree with its leaves, etc.). If an old value of the dimension has been modified and if this value also effects many of the records of the fact table, than the entire indexes/aggregations of the measures get dropped and the processindex takes lot of time (that is not the case with any other "normal" dimensions).
I have found some thresholds in the config file that probably could block the drop of all of the indexes/aggregations, though I am afraid to modify them (they are still on the default setting).
AggIndexBuildThreshold and IndexBuildThreshold
Do you have any experience with them? Would you, guys, suggest to set these thresholds on a higher value or to null not to drop the indexes/aggregations? If they wouldnt be dropped would be the cube still giving back correct numbers after some old value in the dimension has been changed?
Thanks in advance for your answer!
additional info: sql server 2008 standard x64, win server 2008