differential cube building 1

[gbraden]

We are currently building a cube in our environment that is taking about 12 hours to complete. Initially (way before my time) the cube was built using differential data, which I am assuming is the use of new data after the initial cube build for updating.

Searching about on the internet, I am not having any luck finding information about this technique. Has anyone been using differential cube builds? Can you point me to a link?

Baring that, is there any best practices or tune-up advice to get this thing moving just a bit faster? We have already gave the server a lot more memory. I am running SS 2005.

Thanks in advance!




Glen Braden

 

[MDXer]

Look at partitioning your cube. You can partition along any of your dimensions but time is usually the norm. This allows you to process only the data represented in that partition. If you partition by month you only ever process a months data unless you do a full process of the cube. You can also do an update process on yuor dims an facts but becareful of doing it on your dims as it drops the aggregations.

Paul
---------------------------------------
Shoot Me! Shoot Me NOW!!!
- Daffy Duck

 

[gbraden]

Thanks MDXer. That makes a lot of sense. Can you point me to website/knowledge base explaining the use of partitions?


Glen Braden

 

[gbraden]

Found several articles. Here is one, for anyone who might be interested...

http://www.databasejournal.com/features/mssql/article.php/10894_3419631_2

Glen Braden

 

[MDXer]

Project Real

Project real is MS best practices. It is the Barnes and Noble solution implemented by MS they have a number of white papers that will probably give you the true answers you need for this element of your solution along with information for other areas you may not have even thought of yet.

Paul
---------------------------------------
Shoot Me! Shoot Me NOW!!!
- Daffy Duck