RDBMS: Reports of Its Death Exaggerated
February 14, 2009
Tony Bain’s “Is the Relational Database Doomed?” is an interesting article. You can find it in ReadWriteWeb.com here. The point of the article is that “new” databases are becoming more widely available. Relational databases are showing their age, particularly in the petabyte and exabyte world in which we swim. The multi-part article provides some background, a useful table of comparison between the RDBMS and the “new” Key Value Database. The wrap up is a discussion of the upside and downside of the key value database and then a snapshot of new database offerings such as Google AppEngine Data Store, SQL Data Services, and some newcomers like Mongo and Drizzle. I liked the write up. My focus is not on a database replacement. I am more interested in the data management issues and dataspaces. Sue Feldman and I wrote a report on these topics in September 2008. Looking at a database is like understanding an automobile by looking at a dealer’s promotional literature for a vehicle. Useful certainly. Where the action is? Nope.
Stephen Arnold, February 14, 2009
Comments
One Response to “RDBMS: Reports of Its Death Exaggerated”
-
- Subscribe to Beyond Search
Feature archive
News archive
-
Hi Stephen,
Thanks for highlighting this article. As I always like to say infrastructure is un-interesting right up until it prevents you from doing something you want to — then it becomes critical.
As someone who’s been working with and around relational databases since 1983, I believe that the RDBMS has a had a great run, but in time it will be superseded by a plethora of special-purpose databases each designed for a specific purpose and each better than the RDBMS at that purpose (e.g., streams, data warehouses, parallel query processing, text, XML, etc). In fact, I’m somewhat a believer in Mike Stonebraker’s argument that when each given purpose is better accomplished in a special-purpose engine, that there remains no need for a general purpose (jack of all trades, master of none) DBMS as we have in the past. Controversial, I know. But worth thinking about.
As you know, we are a maker one such special-purpose DBMS: MarkLogic Server, optimized for XML content (and lots of it). Other examples: Streambase for streams, Vertica for data warehouses, and MapReduce/Hadoop (the key-value) pairs which I call “un-databases” for parallel processing of simple queries.