I just read a great blog piece by legendary data guru Curt Monash, and it’s titled Why I’m so forwarding-leaning about Hadoop features. The bottom line: Monash is encouraging Hadoop uses to err on the side of new features, and not worry so much about using earlier, more stable versions of Hadoop. His reasoning is that, at this stage in the Hadoop product life cycle, the benefit provided by the latest features is greater than any benefit to product stability.
It’s good logic, of course, we expect no less from Monash. He’s been publishing a number of excellent common-sense posts about Hadoop in the last few years – see Enterprise-ready Hadoop, including its link to Annoying Hadoop marketing themes that deserve to be ignored. Curt’s a great voice of reason amid a lot of hype and bravado.
Curt first got my attention with his discussion about Google’s patent application for MapReduce, and the fact that Hadoop functionality is hardly new, and has long been doable in an Oracle environment. His article titled More patent nonsense – Google MapReduce should be required reading for all data professionals with an interest in Hadoop solutions.
The bottom line: the “big data” trend has merit, particularly for some specific, unique business cases, but there’s a tremendous amount of hype around them, and even Hadoop, at its core, depends on relational logic in order to be of any use.