If you split the cube into several partions by using dates, AS2005 will use more than one processor.
Regards
/Thomas
|||Based on 8 processors I used 8 partitions and it did better, but not much. Then I tried 16 partitions (many programs use 4X threads per processor to maximize parellelization, but if something is well written 1x should be sufficient) and I think it helped - utilization is up to 30 or 40 percent during the cube process part, but the aggregation part is still pretty low - 20-30% overall.
How many procs do you have and what utilization are you seeing during agg processing?
|||The cubes I work with are small and they run on a four CPU box. By using partitions I was able to reduce the the processing time by 50 percent for one of the cubes.
I have not measured utilization, only processing time.
You will have to consider your I/O system as well and memory setting for SSAS2005. Maybe your low utilization depends on bottlenecks in memory or I/O?
Regards
Thomas
No comments:
Post a Comment