Martyn Richard Jones
Gif sur Yvette, 15th December 2017
If all we had right now was operational tech stacks consisting of PL/1 and IMS and Hadoop we would be in a bad place if it came to strategic, tactical and operational decision support.
Not because of PL/1 and IMS but because of the fact that Hadoop is shit at doing many things.
Hadoop heaves super-sized, fast and variable unstructured data – at a respectable, if not fast, pace. That’s what it is good at.
If relational databases were new, we’d be clamouring to replace the Hadoop pond-life with relational and object-relational databases – e.g. SQL/Server, DB2, PostgreSQL, Oracle, etc.
So, why do we want to be so perverse?
Hadoop is not post-modern, it’s pre-classic. It’s so old it’s archaic. The IT of jungle law, anti-intellectuals and bottom-feeders.
Thanks for reading.
Of course, relational databases are based on a sound theoretical model i.e. The Relational Model of data.
This is a sound, mathematical model defined by Ted Codd in 1969/70. The model was then elaborated by Ted and Chris Date and others in the 70s and 80s. Vendors started to implement RM in usable form and the rest is history.
Relational is now the lingua franca of data covering off 95% of application types yet has been under appreciated by many in the industry and relegated to an after thought in the great HADOOP pseudo debate.
All power to your elbow Martyn.