Despite the headlines, traditional SQL databases are not yet dead, writes Wikibon Analyst George Gilbert. But it is “beginning to bleed from a thousand cuts” as new Systems of Intelligence demand new database types with different capabilities.

The traditional RDBMS that has dominated database technology for 30 years is designed for systems of record, which are those that control record-keeping and operations. While RDBMS is very good at what it does, it is designed principally for stability because chance breaks things. The vendor needs to anticipate all needs, and that’s why SAP has about 100,000 predesigned tables. But if users need to capture information that doesn’t fit into those tables they are out of luck.

The problem is that new Systems of Intelligence, unlike systems of record, are continually evolving and need to keep track of constantly changing information. One example is user profiles for consumer Internet services. To continually get smarter about customers, these services need to collect information about them continually, some of which is impossible to predict in advance.

New workloads require databases that can do a small number of things very well, easily and cheaply. Serving ads to online users at the point of the purchase decision, for instance, demands extremely fast databases that can identify specific patterns of behavior. Big Data requires an entirely different kind of database, able to manage trillions of rows and millions of columns. Gilbert identifies 10 kinds of database, each matched to a different set of needs. And yes, traditional SQL databases such as Teradata are in that list, but they only are appropriate for a specific set of functions.

Read the full report on Wikibon Premium.

photo credit: CallieDel Boa via photopin cc