We have a data mart that contains data going back 6 to 7 years. Up until now we usually roll off the earliest year. Users have requested not to roll off data any longer as well as going back earlier in time and adding more data. The older data is static. Currently our data mart process is a drop and replace. We use all surrogate keys.
My question is what are the thoughts for the best method for not having to refresh the fact and dimension tables for that older static data so as to keep the build time manageable.
Sorry in advance for the newbie methods we have thought about implementing. One idea was to build the DM as normal with all the required data to satisfy user requirements. We thought about finding the max Identity (PK) in each fact and dimension for records that would remain static and also set a flag on those records. Then each month delete the years of data that would need to get refreshed and perhaps reseed the table based on the highest value of the static data. Then just process the newer set of data and appending to the fact and dimensions.
We are probably way off base here. So please excuse the ignorance on our part.