JavaZone 2009 - The Forgotten Art of Batch Programming
You may think this is a stupid presentation. Who would ever get millions of transactions in files from a legacy system? Transactions will be placed from Web and properly load balanced as they get in. But the fact of the matter is, don't be so sure. There is a lot of legacy systems out there. In 1997 the Gartner Group estimated that there were 240 billion lines of Cobol code in active apps. Something like 90 percent of financial transactions are processed by Cobol code, and 75 percent of all business data processing is Cobol. Most of that code is still there after being fixed for 2000 bug.
So why is all this Cobol stuff relevant to me? I develop using Java, not Cobol (Thank God!). It's relevant because Java is ready for the heavy lifting of replacing all this code. But since all systems can't be replaced in one go, you must interface with the existing solutions. Doing that using typical Java stack in the typical Java way, can give you an over-engineered and badly designed solution.
The speaker spent the last 3,5 years as a software architect and developer replacing large mainframe applications using best practice Java architecture (Message oriented, Spring, Hibernate etc). Doing so he realized that new techniques have their place, but processing large batch files using Domain Driven Design and Hibernate probably wasn't the easiest way of dealing with legacy code and integration.
Morten Udnæs
Morten currently works a senior consultant for Miles AS. His has been an active developer since the early 90s developing financial applications using Cobol/Mainframes, Client-Server technology, Microsoft .NET and Java. Learning the hard way that complexity is always bad, he spends his time getting better at using Agile methods, lightweight architectures and Cloud Computing.