My friend and fellow blogger Steve Overton was tasked with presenting not one by two papers in a single one hour block. My original strategy was to take Instagram pictures of him running down the hall to get to the second session on time, but since both were in the same room (darn) my new plan was to actually learn something. And wow, Steve presented two great ways to address data administration.
431-2012 "10,000 Leagues of Data... Divide and Conquer OLAP Cubes" presents a interesting way to leverage the incremental update capabilities of SAS OLAP by adding sections of the data into the cube one at a time rather than inserting millions of records in one swoop. This avoids any possible server resource constraints of loading super large cubes. Dealing with lots of data, ROLAP reduced speed for the end user while aggregations caused temp space failures. Steve created a script that takes care of dividing the data into multiple tables & running the cube on each data slice. The only trick is to create a single metadata table definition since this is required for PROC OLAP to run successfully.
In 112-2012 "Developing a Flexible ETL Process to Let SAS® Capture Data Changes ~ Efficiently in a Data Warehouse and Clean Up the Mess", Steve has created a way for SAS to manage SAS data table changes by adding a control table to the ETL process. A customer's project implementation required this method because of different data update schedules and only changed data should be processed incrementally. Here is a high level diagram, grabbed from his paper, on how this works conceptually.
This is just one more reason why I so enjoy attending SAS Global Forum. Everyone has a solution or idea that can be used elsewhere to improve. We are all focused not just on sharing our knowledge but learning from others to continually improve.
So, what have you learned today at #SASGF12?