NuoDB, ParElastic Racing to Build Database Systems of the Future
It ain’t exactly Microsoft vs. Oracle circa 2000, but the Boston area has its own little database war going on.
I’m talking about NuoDB vs. ParElastic. What, you haven’t heard of either one? Well, that’s because they’re working on the world’s next big database system. And you probably haven’t heard of that either, because it doesn’t fully exist yet.
OK, this is all just a subset of what’s happening in database technology—and, more broadly, in the infrastructure subsector of “big data”—but it’s a pretty important problem these companies are trying to solve. (You can catch NuoDB and ParElastic on the same panel at our “Future of Big Data” conference in Boston tomorrow.)
For decades, the relational database technology of giants like IBM, Oracle, and Microsoft has been dominant in industry because it’s been good enough. But now the tide seems to be turning. Data centers and virtualization systems have become commoditized, so essentially there are lots of small machines crunching information rather than a few big mainframes.
Plus, as you have probably heard, the amount of data being collected from networks, sensors, mobile devices, and the Web has skyrocketed. And what a database has to do in 2012—handle hundreds of thousands of users, queries, and transactions quickly—is vastly different from what most were designed to do.
Most database companies offer tradeoffs, says NuoDB CEO and co-founder Barry Morris. If you want to handle tons of queries fast, you can’t handle transactions. You can read data quickly, but you’ll write it slowly. You can get business intelligence insights and analytics from historical data, but you can’t run your business day-to-day with the same system.
Unless you try something radically different, that is. NuoDB co-founder Jim Starkey, a DEC veteran and longtime database techie, took a fresh look at the problem and saw that the databases of the past 30 years were built in an obvious way that was very successful, but doesn’t scale for today’s usage. “Jim asked, ‘Is there a non-obvious way?’” Morris says.
Morris, the former CEO of IONA Technologies and StreamBase Systems, was impressed after seeing Starkey’s ideas. So were venture capitalists Morgenthaler Ventures, Hummer Winblad Venture Partners, and Longworth Venture Partners, who have put some $20 million into the two-year-old Cambridge, MA-based startup. “We’re saying you don’t have to give up anything,” Morris says, referring to scalability and performance. “I said if this thing works, it’s the future. It’s the 21st century database.”
NuoDB’s technology is an “emergent” database. Instead of having a central control point or hierarchy like a traditional database, it has smart objects, or “atoms,” that work peer-to-peer. Some of the atoms effectively index the data, some describe where other atoms are, and so forth. There are about a dozen kinds of atoms, Morris says, and together they manage and represent all the data in the database.
“It’s the only way that nature builds things at scale,” he says, drawing an analogy to animal migrations and flocks of birds. “A few simple rules result in macro intelligence.”
That all sounds pretty ambitious—researchers have been working on swarm intelligence and emergent systems for decades—but the key is that to the outside world (and customers), the database behaves like any other, but it also scales simply by adding machines.
NuoDB has been testing its system with beta customers and is looking to do a general rollout later this year, targeting both Global 1000 companies and startups trying to build the next Facebook. “I’m here to tell you we’ve got it, and it’s working,” Morris says.
In terms of the business model, he calls NuoDB the “iPad of databases,” meaning it’s intuitive to use and easy to get started online. “I don’t think we’re going to sell this by explaining all the intricate and boring technical details,” he says. “It’s ‘Try it, you’re going to have your socks blown off.’ It’s much more Apple than it is Google, let alone IBM.”
Meanwhile, ParElastic, based in Waltham, MA, takes a very different approach to the same basic problem. “We’re trying to take advantage of the technologies that are already out there,” says CEO and co-founder Ken Rugg. He and co-founder Amrith Kumar are veterans of Progress Software and Netezza, respectively, so they know a little something about the database industry.
As Rugg explains, “The transformative change in the industry isn’t a new database technology, it’s this transition in the business model. People will stop purchasing databases on servers, and start purchasing databases as a service.” And ParElastic is trying to build the technology that enables that transition.
Without getting into the nitty-gritty details, the two-year-old startup has software that sits between companies’ applications and the underlying (and existing) databases. The software enhances the capabilities of databases—making them more flexible and “elastic”—by harnessing multiple database servers and making them behave as one server from the application’s point of view.
The secret lies in how the software manages all the different loads and resources so that business customers get all they can out of their data systems, while minimizing any disruption to existing IT architectures.
ParElastic has raised just under $3 million from the likes of CommonAngels, General Catalyst Partners, Point Judith Capital, and LaunchCapital. The company is currently running beta trials with its customers.
It’s still very early, of course, but Rugg doesn’t dispute the notion that ParElastic could become something like the next Oracle if all goes well. He does emphasize, however, that his company is “not trying to build a faster database, we’re trying to build infrastructure” to help virtualize databases in the cloud.
So who wins in this sector? Both startups in question have patents that were granted quickly—NuoDB for its emergent database system, and ParElastic for its database management architecture. NuoDB sounds like it’s trying to do something radically different, which is higher-risk but potentially transformative, whereas ParElastic is using more traditional methods to extend the usefulness of existing data systems.
In the end, I doubt that this sector is winner-take-all. It’s more likely that these companies, and many others, will be part of an overall transformation of enterprise IT systems away from the IBMs and Oracles, and toward sleeker, nimbler architectures. But that could still take quite a while.