Santa Cruz, Calif. -- Hitachi Ltd. may be a giant company, but Malcolm Colton, vice president of sales and marketing at Hitachi America's Embedded Business Group, is starting from scratch with a new message. His group was recently formed to market Entier, a new relational-database management system for embedded devices.

Hitachi has decades of experience with relational-database technology, and in Japan it competes with database giants such as Oracle. But Hitachi's data-management products have had "zero footprint" outside Japan, Colton said. Instead of trying to increase its global reach in a relatively static enterprise market, he said, Hitachi's leadership decided to try something different.

"They noticed that devices are getting to be powerful enough, and to manage complex enough data, as to require data management," Colton said. "So they built a brand-new data manager for the device market that fits roughly into a half megabyte of space. It's not a cut-down version of an old product; it's a brand-new product for a brand-new market."

There are more embedded devices than PCs these days, Colton said. "The average iPod has more storage than my laptop, and people are starting to need sophisticated searches on these platforms," he said.

Relational databases traditionally consume hundreds of megabytes and serve large, enterprise-based applications. So why bring them down to the embedded-device level? The primary reason, said Colton, is to provide local search capabilities. For example, said Colton, people might want to find music that fits a particular event or mood. A field worker might want to do a spatial search to locate jobs in a neighborhood. Emergency services personnel may need to find local information in a hurry, even though they're disconnected from a central network.

Better at search
Object databases such as db4o (see Nov. 20, page 22) are also making a pitch for embedded devices, promising simplicity and efficiency. But relational databases offer much better search capabilities, Colton said. "We expect to see some penetration of object databases for people who are persisting Java objects, but if they want to do search, they'll want a relational database," he said.

Colton believes he can quickly deflect any claims that relational data managers take up too much memory. Entier can provide a reliable file store in only 90 kbytes, he said. A minimum-size relational database with SQL search capability will take 300 kbytes. Adding special capabilities of Entier, like an incremental text search and a spatial search, will take another 300 kbytes, he said.

There are no standard benchmarks for speed, Colton said, but he pointed to a demo on a PDA in which a user can do a music search in about 10 milliseconds. Entier also offers fault-tolerance, he said, and it performs automatic recovery.

But where Entier really shines, according to Colton, is in search. The data manager offers an incremental text search that progressively executes the search, narrowing down the results with each keystroke and guiding the user with a list of possible next letters in the search term. This can be used for applications that involve searching by name.

There's also a spatial, SQL-based search that works with points, circles, polygons and routes. This makes it possible to bring location-based services to mobile devices such as navigation systems and location-aware smart phones.

The Embedded Business Group offers the Entier software development kit, training and services. Most revenue will come from run-time royalty licenses. Colton said Entier works with many real-time operating systems and recently completed a port to Wind River's VxWorks.

In addition to object databases, Entier will compete with "lightweight" databases from relational-database companies, and SQLite, an open-source embedded data manager.

Hitachi America's Embedded Business Group is starting out with seven employees in Santa Clara, Calif. Development work is done in Yokohama, Japan. Most of the prospective clients for Entier are not Hitachi customers. Will name-brand recognition help? "It's too soon to tell," Colton said.