CODEWITHSUNDEEP
sql-server-2008ssisssascubemeasures
user2889510
user2889510

Reputation: 11

Dimensions and MeasureGroups Prcessing

I have 10 dimensions and 10 measuregroups-- each measuregroup is having one partition each.

All the cube when processed it takes 5-6 hrs daily. so i thought to process the cube using SSIS analysis Services processing tasks where 1st i process dimensions "Process Full" and then i Process all the measure groups "Process Full".

MeasureGroup-1 analysis Services Processing task alone which has only one measuregroup takes 2 hours and rest of the 9 measuresgroups are in measuregroup-2 analysis Services Processing task take 2 hours.. So totally it will take around 3 hours to complete processing rather than 5-6 hours.

So do u think this approach is correct!! Please Suggest..if any constraints/Dependancies are there which Provides fault data after processing like this.

Upvotes: 1

Views: 343

Answers (1)

FrankPl
FrankPl

Reputation: 13315

You should measure the amount of time needed. In many cases, doing all in one go can be faster than splitting the processing, as some data can be kept in memory from dimension processing, and the writing to disk need not have finished while measure group processing runs. On the other hand, your approach may be faster.

In general, the only correct answer is: run a test, as there are many factors influencing this.

To optimize processing performance, you would have to find the bottlenecks: Maybe more RAM would help (do you see that most of your RAM is used during processing? Then, testing a server with more RAM may be worth a try); maybe more CPUs allowing more parallelizing would help; maybe optimizing the relational source so that the statements that Analysis Services issues run faster would help; probably reducing aggregations would help to improve processing performance, but might compromise query performance; moving the AS storage to a different disk from the relational data (if they so far share a disk) should improve the performance; faster disks may also be an option worth testing.

Like many performance optimization tasks, without knowing the details, and running some tests, it is difficult to say which changes would improve or worsen the situation.

Upvotes: 1

Related Questions