Recent developments have reinforced the need for change in our political system. In the same vein the blog post below provides a path to unlocking your data and compute from mainframe. The content in this post is a straight ripoff from my buddy and colleague David Malone's postulation on this topic. The steps below from Dave and are the most cogent explanation of how to move data away from mainframe to distributed systems. It draw's from Dave's direct experience at a major retailer. The same approach can be leveraged to starve and strangle data locked in any data store.
A strangler approach to modernizing COBOL CICS on AS/400 insurance claim processing apps:
1. Build a webservice in front of it
2. Setup change data capture (CDC) or other similar approach to propagate data from DB2 to distributed
3. Write new webservice, point all new systems to it
4. Create sync / dedupe process to move data from distributed back to DB2
5. Identify all existing integrations to DB2 that are performing writes
6. Establish a plan to remove those write systems, and do it. This is where the abstract and replace pattern using technologies like Apache Geode can help.
7. Once write systems are removed, cut off CDC integration
8. Long-term, as you touch the remaining read systems dependent on that DB2 system, reinterface them, or remove them if possible
9. Once all remaining read based systems are remediated, shut down backwards sync, delete DB2 instance
A take on how to modernize compute processing in mainframes is covered in a session from SpringOne2016 - Get Off The Bus Gus - 50 Ways To Leave Your Mainframe. This presentation talks to moving batch processing off the mainframe.
Thanks to Colin Stevenson for triggering this conversation on slack