NoSQL Rebels Aim Missile at Larry Ellison's Yacht

In Silicon Valley and beyond, a new kind of database is rising. Dubbed "NoSQL" by its proponents, it sprang out of Google, Amazon, Facebook, and other web giants, who used it to run their massive online operations. But now it's moving into the rest the world, backed by a growing number of startups.

Unlike traditional relational databases from the likes of Oracle, Microsoft, and IBM, NoSQL databases are specifically designed to store massive amounts of data across an enormous number of servers. They don't give you quite as much control over the data as a relational database, but because they don't follow fixed schema, they can also provide an added degree of flexibility -- and they can get you around the hefty licensing fees charged by the Oracles and Microsofts.

According to analyst firm The 451 Group, the market for NoSQL databases is growing at a torrid 82 percent compound annual growth rate, with a startup called 10gen pushing the open source MongoDB database, the Texas-based DataStax putting its weight behind the Cassandra database developed at Facebook, and others backing software such as CouchDB and Hbase. Increasingly, traditional enterprises are learning that NoSQL databases can actually handle the majority of even traditional database workloads.

The big guys are doing their best to suppress this uprising, but they're vulnerable -- none more so than Oracle.

It's not time for Oracle CEO Larry Ellison to sell his yacht. At least not yet. Though leading NoSQL database MongoDB is the second most in-demand technology skill on Indeed.com, it and every other NoSQL technology is a comparative pygmy when stacked against the demand for Oracle relational database management system (RDBMS) and Microsoft SQL Server skills.

But if you look at relative growth in demand, Oracle and SQL Server are rounding errors, as is yesterday's heir apparent to the SQL throne, MySQL.

Ellison isn't one to get rattled easily. After all, this is the same guy that pooh-poohed cloud computing for years and has deprecated threats from infrastructure partners like Red Hat. When you've got billions to spend, you can buy relevance even when you can't build it.

That's the hope, anyway, as Oracle certainly seems incapable of building a winning NoSQL strategy. Though you've probably never heard of it, Oracle does, in fact, have its own NoSQL database, which it says is great for online display advertising and mobile social gaming. Unfortunately, Oracle apparently can't point to a single customer in either segment for its NoSQL product. This might be because Oracle presents (PDF) its NoSQL technology as inextricably linked to its more expensive RDBMS: "Complementing the Oracle NoSQL architecture is the Oracle RDBMS that is critical to the overall solution."

How comforting.

So perhaps Oracle will eventually buy a leading NoSQL vendor and hope to corral its community, the way the company has done with MySQL, with largely positive results. In the meantime, however, Oracle more than most is essentially locked out of the NoSQL game in ways that IBM and Microsoft aren't, with long-term negative effects for its core RDBMS business.

Microsoft, after all, has been actively partnering with NoSQL upstarts like 10gen, not to mention talking up the technology with developers. IBM, for its part, not only is doing its own Oracle-esque blocking-and-tackling of NoSQL by incorporating some of its functionality into DB2 in a nod to developers, but has the benefit of being fueled in large part by services dollars, which services can be around its own DB2 technology or third-party NoSQL technology.

Oracle lacks such a services arm, and it has so much invested in SQL through its core database business, along with its acquisitions of BerkeleyDB and MySQL, that it will be difficult for the Redwood Shores giant to truly embrace NoSQL.

Meanwhile, in textbook Innovator's Dilemma fashion, NoSQL has started off as optimized for web applications but is quickly becoming more than good enough for a wide array of enterprise applications that have traditionally been built with relational databases. NoSQL databases are unlikely to handle complex transactional workloads in the near future, and VMware's Dave McCrory suggested to me that a traditional RDBMS will always be required in 15 to 20 percent of workloads.

But this still leaves the vast majority of database workloads open to NoSQL cannibalization.

So what happens when an enterprise discovers that open-source and free Cassandra or CouchDB or Riak or MongoDB, brought in to handle an intranet application or an commerce system, is also good enough (and dramatically cheaper than Oracle's RDBMS) for a few projects slated for Oracle? As 10gen president Max Schireson notes of his company's MongoDB customers, they're almost certainly going to buy more MongoDB. Lots more:

"I want to save a customer millions of dollars and charge them a modest fee," Schireson says. "Why? Because when that happens they’ll be aggressively looking for the next place to use MongoDB. They will tell their friends not just about how great the product is, but how easy 10gen is to deal with and what great value we provide. Short term revenues may be less, but if this makes the business grow faster over time revenues will be much higher."

Given that roughly 60 percent of the world's databases are operational in nature, this will happen all the time, even in the traditional, stodgy enterprise. There are many applications currently reliant on relational databases that would be better served by NoSQL databases. Soon they will be.

This will start to accelerate given that developers increasingly drive enterprise technology decisions, and those developers are opting for NoSQL. Oracle may own yesterday's database administrators, but it's losing tomorrow's database developers.

This same trend impacts Microsoft and IBM, but both are hedging much more effectively than Oracle. And while I'm not expecting Oracle's revenues to drop off a cliff, especially given the long-term maintenance contracts it has in place, NoSQL is cutting Oracle out of the future of enterprise computing. If that isn't keeping Ellison up at night, it should.