How to handle historical data change in SSAS

Question

How to handle historical data change in SSAS

PKDC

SSChasing Mays

Points: 618
More actions
February 15, 2013 at 9:07 am

#150210

I am a novice to SSAS and I was wondering if there is a way to process historical data change in a fact table? Time to time we do need to make adjustment to the data that was loaded to warehouse. In stead of doing full process, is there a way for SSAS to recognize the only updated info and process? Any help on this would be appreciated.

Viewing 6 posts - 1 through 5 (of 5 total)

You must be logged in to reply to this topic. Login to reply

Ells SSCrazy Eights Points: 9901 More actions · Answer 1

What is the database that you are looking at. If SQL Server then possibly CDC(change data capture) could be of interest? That would identify the changes.

Then process update could avoid a full process (depending on what changed).

E

PKDC SSChasing Mays Points: 618 More actions · Answer 2

We have a process to identify the changes, but I am more curious about how to process incrementally if the data is altered was 2 weeks old data or older(cloud be any historical date range) entry in fact table. Any help would be appreciated.

Ells SSCrazy Eights Points: 9901 More actions · Answer 3

This location documents the processing options for Analysis Services objects. Have a good read through and see what will work for you. If it is historical data how regularly are you going to update it? Would splitting the cube into different partitions help?

E

RonKyle SSC-Dedicated Points: 31554 More actions · Answer 4

It sounds like your fact table is an accumulating snapshot. So snless the cube is split into partitions, and you can be certain that all the changes are in certain partitions, you will have to do a full process. Is the time so long that this is unrealistic. Mine have millions of rows and compile in about 10 minutes, but I can easily imagine that this might be small compared to others.

MMartin1 One Orange Chip Points: 27794 More actions · Answer 5

How you define your partition is partly if not fully influenced by business logic. If a sale has a 30 day return period then you can create each partition in monthly increments more or less. You just process the affected partition rather than the entire cube.

I think the process update only works on dimensions, or am I wrong?

thanks.

----------------------------------------------------