Just when we’re getting all cozy with relational database systems, one pioneer of the industry is saying that the relational database is going to go the way of the dodo bird. Professor Michael Stonebraker (the prime mover behind industry standard relational databases Ingres and Postgres) has a vested interest – a startup to promote – but the idea is sounding very timely.
Stonebraker’s new company, Vertica, is based (as you might have guessed from the name already) on the notion of storing like data with like: in columns, if you will, of structurally similar data, rather than rows of structurally dissimilar, but related, data. As I understand it, this means you’d have a column of phone numbers, another column of names, another column of addresses, etc., each stored in a discrete data set, rather than a single table with phone numbers, names, and addresses all aggregated together into a table of addresses.
In a way, it seems akin to the mashups idea that’s creating so many great new web services – instead of pulling together massive amounts of data into a single physical/hardware space, thinking in columns vastly opens up the possibilities of distributed data storage, bringing together disparate pieces of information on the fly into record “mashups”, in real time, as they are needed.
That’s the best explanation that I can come up with on my own … if this intrigues you, take a look at Stonebraker’s blog site, because I’d be curious to hear what others think.
Update: I trolled around the Vertica website for a bit, and noticed a reference to RDF data management. Sure enough, there’s a whole page on using Vertica for RDF-based Semantic Web applications.
