Suppose I have 2 datawarehouses of two different companies that are being merged. Some of the tables are common but includes different records, while others are different. If I use big data environments, can I put all the tables together as data swamp and to create new data lake with archive of the previous datawarehouses data or should I build a new datawarehouse based on the common tables and adding the different ones? In addition, is there a good source for best practice steps for merging datawarehouses?
There is a design pattern call "strangler" that might be applicable. The strangler design pattern leaves all legacy systems in place and migrates piece-by-piece to a single, updated system. It does this by creating a proxy interface that routes requests to either legacy system or the updated system. As the migration happens, the proxy routes more traffic to the updated system.
The advantages of the strangler design pattern:
- People can still use legacy system
- Prioritize system clean-up based on usefulness
- Can create an explicit roadmap of migration