If you read a lot of blogs and IT industry commentary online, at some point this year you will have come across an article from someone talking about the ‘death of the RDBMS’. Even though RDBMSs from the major vendors (Oracle, IBM, Microsoft) are still very much a corner stone to most large-scale business systems today, you can’t help but notice that a lot of new systems being built that address problems of enabling massive online services are not using RDBMSs any more. Huh? No, they’re using other data storage mechanisms that are turning their back on the past 30 years or so of experience built up from using RDBMSs, or is it that we’re actually reaching the point where we’re realizing maybe RDBMSs are not actually the best/most appropriate way of storing and retrieving data for all scenarios?
Most of the alternatives to RDBMSs that seem to be gaining traction seem to have roots in Google’s approach of BigTable and Google File System, where information is stored in a heavily denormalized format with tables having thousands of columns. Rather than breaking data down into relationships an separate tables, the main feature of this approach is simply representing data as key/value pairs. You have a key, you have a data value. Simple.