August 9, 2006 at 5:34 am
im trying to cut down on the amount of data accessable from within the cubes, as some of the
pc's in my workplace are old and have a hard time retrieving 4 years + of data.
my idea is this:
i am going to archive data in the time dimensions, so that only the last months are left in
the time dimension table.
what i want to know is this, do i have to also archive the fact data associated with this
archived dimension data? it wont apper in the cube, if it has no time dimension related to it
i would think . there are many other dimensions though, so maby the old data could get brought
into the cube by association with the other dimensions. is that possible?
August 14, 2006 at 3:50 am
If I am reading this correctly I suspect that you may have a problem here and you will need to archive the fact data associated with the months that you are archiving.
Assumming that you have a foreign key relationship between the fact table and the time dimension you will not be able to delete a parent record where child records exist. If you disable the constraint to allow the delete the cube build will then fail (assuming molap) as it will look for the parent record(s) when processing the cube.
Essentially if data is in the fact table it will be processed and will be in the cube
August 14, 2006 at 8:13 am
Is there any rerason why you can't create a view to retrieve only the relevant records (from time and fact) and set these views as the source/s for the fact and dimensions?
Alternatively, you could create a specific 'current data' cube for those users who are having issues bringing the data to their machines. On that topic, you mention that they can't bring the data to their machines - have you got any aggregations set on the cube? In general, not 'all' of the data is sent to the client (speaking in AS2000 terms), only the metadata required to create queries and then the data returned by queries. You could have 50 years worth of data but if they run a query that affectively says return year on rows and sales amount on columns, thats only 50 rows or 101 cells (labels plus values) returned to the client. Are your users receiving error messages or somethign else that is causing you to think it's the data volumns? What tool/s are they using?
Cheers,
Steve.
Viewing 3 posts - 1 through 2 (of 2 total)
You must be logged in to reply to this topic. Login to reply