Riding The Clustering Wave: Tsunami Research

Tsunami Research's hive computing is a cluster technology that puts the smarts in software

In the computer industry, the right metaphor counts. Consider the term client-server computing: It's simple, appropriate, easily understandable. Not so Web services, which has suffered a good deal of confusion, in part because of its unwieldy designator. While Web services is likely to overcome that, the list of failed metaphors is long: artificial intelligence, expert systems, and neural networks, to name a few.

The latest computing metaphor to grab the popular imagination is grid computing. It refers to a network of computers, from mainframes to idle desktops, lashed together to bring collective computing power to bear on a single, usually computationally intensive, problem.

Now comes hive computing. The metaphor comes from bees, and the fact that any bee in the hive can be called on to do any job. Tsunami Research Inc. in St. Louis, which develops and markets hive-computing technology, was started by folks who previously developed a proprietary payment-processing system.

Like grid computing, hive computing leverages a collection of commodity hardware. Unlike grid computing, hive is meant for business-style applications, such as transaction processing and database access. Tsunami boldly claims its technology can lower acquisition costs by 90% and software development and maintenance costs by 50%.Tsunami's flagship toolset, HiveCreator, lets developers write applications that are self-organizing, self- maintaining, and self-healing, according to the company. That means software gets the hardware to work together, keeps the app up and running, and recovers it when crashes occur. Tsunami says hive computing is more scalable and can recover from crashes more easily than other cluster-computing approaches.

"The model is, instead of investing aggressively in hardware, put all the smarts in the software," says Mary Johnston Turner, VP of enterprise strategies at research firm Summit Strategies. "It turns what's been going on with virtualization and grid computing on its head."

One limitation: Hive computing can't be retrofitted to existing applications. "You really have to write the application from the ground up," Turner says. Tsunami says it has several customers testing the product, most in industries that write their own applications, such and finance or telecommunications.