Is the relational database model dead?

Are vendor claims that big data, IoT and NoSQL killed RDBMS credible? Nothing is that simple in the enterprise data center.

Is the relational database model dead?

I recently received a promotional message from a PR representative of a supplier of database technology trying to arrange a briefing prior to the upcoming Oracle OpenWorld. The come-on was based upon the notion that the needs of big data, Internet of Things (IoT) and the tremendous growth of smartphones, tablets and other intelligent devices have overcome the capabilities offered by relational database engines, such as those offered by Oracle.

+ Also on Network World: Does 'share-nothing' NoSQL signal the end to system resources sharing? +

The PR rep's goal, of course, was to arrange a conversation between an executive of a supplier offering a non-relational/NoSQL database and one of my analysts or me. The PR rep said, "The heyday of the 30-year-old relational database is over, and Oracle has missed the shift to next-generational database technology needed to meet today’s big data challenges" in the hopes of persuading me to invest an hour to speak with the executive. As it so happens, I know this executive and we've spoken from time to time over the past 15 or 20 years so—a more direct approach would have worked.

Is this creative approach based upon a wrong assumption?

While the approach was most certainly creative and got my attention, it is clearly based upon an incorrect assumption—enterprises have the ability to easily and quickly discard technology in use since the late 1960s and pivot to something new. I’m sorry to tell this vendor that enterprises simply don't change their spots that quickly.

Adoption takes a long time and so does abandoning technology

In my experience, it takes a long time for new technology to make its way from the development lab to research institutions to academic institutions to appear in enterprise data centers.

That adoption typically doesn't mean an older approach is abandoned and the machines supporting it are turned off, dismantled and the pieces taken to the loading dock to be shipped to a recycling facility. What really happens is that new technology is tested out in a new application that reaches back to the old systems for support and enterprise data.

For example, a new application, based upon new technology, might be used to gather up data from smartphones, tablets and the like. This new application, in all likelihood, will simply feed the data back to established systems in the enterprise data center. In short, the new technology sits on the shoulders of older technology. That older technology sits on the shoulders of still older technology. If we examine the data center for most large organizations, we'll see systems designed and built in the 1960s supporting tasks designed and built in the 1970s, which are supporting tasks designed and built in the 1980s and so on.

Yes, the enterprise data center is an electronic tower of technology.

There are only a few times that everything supporting a specific application gets moved out of the data center. They are:

  • A business unit has been sold, and all of the applications, data, staff, etc. are transferred to another company. In this case, the technology isn't abandoned, just moved.
  • An old stack of technology supporting a workload is replaced lock, stock and bagel by an entirely new approach. A company's homegrown customer relationship management system might be replaced by a new, cloud-based solution. In this case, the old systems and software might just be reassigned to support another workload.
  • A company goes out of business, and everything is on the auction block.

Even new database technology has roots

Another point is that the roots of today's non-relational or NoSQL databases can be found in earlier technology. There were data management approaches that worked quite well that existed before the relational model become popular.

In my software engineer days, I worked on "navigational" databases constructed out of many individual files and indexes, multi-way balanced tree databases such as MUMPS or Pick, CODSYL databases, or network databases all in the 1970s and early 1980s.

Today's NoSQL database technology didn't spring fully formed out of some software engineer's mind in the 2010s!

Calling BS

I guess that I have to call "BS" on this whole approach for a few reasons, including:

  • Enterprises build upon older technology; they do not throw it away. So, the idea that something new is going to come in and enterprises are going to throw everything away isn't credible.
  • NoSQL and in-memory database technology is an evolution based on technology originally developed in the 1960s and 1970s. So, the supplier in question can't claim what they're doing is strikingly new and original.
  • Oracle has acquired and is selling in-memory and NoSQL database technology alongside of its popular relational database products. So, the supplier in question can't support its claim that Oracle missed an emerging trend. What might be true is that Oracle is facing "the innovators delemmia" and is trying to preserve a big stream of revenue while it transitions to products and technology that currently produce a much smaller stream of revenue.

I look forward to having the conversation with the executive of this supplier to learn how he's going to present his messages.


Copyright © 2016 IDG Communications, Inc.

The 10 most powerful companies in enterprise networking 2022