I have 1 year of data and tried partitioning large measure groups per month. The oblivious reason why we do partitioning is to process only the latest data coming in. But there are instances that we need to process the whole cube/all partitions. The problem that I am encountering now is the processing of partitioned measuregroup using ProcessFull is 100% slower compared to non-partitioned measuregroups. I have already tried processing it sequentially and in parallel. Are there anything I need configure in the server?
Asked
Active
Viewed 195 times
1
ggarcia
- 47
- 1
- 10
-
Couple of things: 1) Try dynamically detecting which partitions actually have changed data in them to reduce the amount of work and processing needed. 2) Check the amount of concurrent connections you have to the source data and maybe increase that to deal with the extra processing. – Paul Andrew Jun 01 '16 at 11:16