Loraine Lawson did a great job in covering Burton's take on busting data silos.
"Data silos are inevitable, according to Lyn Robison of the Burton Group. Silos are a natural part of decentralized IT systems, he argues, existing not just because of decentralized information systems, but because of the departments, groups and teams that are part of everyday, enterprise life."
The core notion is that Burton is recommending that instead of busting these silos through approaches such as SOA, that we learn to accept them and integrate them with other silos within the enterprise. In essence, do the most with the least. I could not agree more, but if you think this is new, you're kidding yourself.
Sometimes I feel like the old programmer in the room that understands perfectly well that most of what's done today within IT is really building upon older concepts. The trouble is that new people come up through the ranks and often overlook, don't know, or don't care about existing approaches that are well understood and proven. The concept of data integration is one of those. Okay, rant over.
The fact of the matter is that integration between data silos has always been an option, even in these days of SOA and cloud computing. Indeed, it's typically the most successful considering that it's much less invasive, and much less expensive than SOA, or more complex integration approaches that require adaptation to the integration solution, rather than the integration solution adapting to the existing problem domain. The former is much more expensive and risky.
Core to this would be the old integration concepts of replication and abstraction.
Replication, is really about moving data between silos accounting of for the semantic differences using some sort of translation and routing layer. You can replicate the data either in batch or near real time.
Abstraction, is about mapping physical databases to virtual logical databases, thus providing integration through data visibility using a virtual, but typically much better designed structure, that's bound to physical databases within silos that are typically not well designed.
Either approach works, but they are really old approaches. Nonetheless, should be considered as primary cost effective and noninvasive approaches. Sometimes less is much more.
Leave a comment