I discerned a couple of recurring themes when I sat down with ParAccel CEO David Ehrlich to discuss the venture-backed startup that specializes in database management systems technology.
One is that ParAccel’s industry seems prone to what Ehrlich calls “database wars,” which might be best exemplified by the intense competition that broke out in the 1990s between Oracle, Sybase, Informix, and others. Another is that the computerized data storage networks that were designed for data warehousing have not proved to be well-suited for data analytics.
Both themes weave through the ParAccel story, which was founded in 2005 by Barry Zane, a co-founder and system architect at Netezza who left the Framingham, MA, company to start ParAccel in San Diego, and who now serves as its chief technology officer. Ehrlich says another key technologist is Rick Glick, ParAccel’s vice president of technology and architecture, who was previously the chief technology officer of the San Diego-based database engineering group at Teradata of Ohio.
Ehrlich, who divides his time between Par Accel’s offices in Cupertino, CA, and San Diego, says the company’s “center of gravity is in San Diego” largely because Zane started here, and both Zane and Glick continue to live in San Diego. Another factor is the emerging cluster of software analytics in San Diego, as well as the presence of Teradata’s engineering group. As Ehrlich puts it: “You just can’t find people anywhere else in the world that have 15 to 20 years of experience in parallel processing database computing.”
ParAccel, which has raised about $53 million in venture capital and now has 70 employees, specializes in relational database software that enables its customers to combine the benefits of data warehousing with analytics processing. Its technology is designed as a column-based database management system that runs on a massively parallel processing hardware platform.
As Ehrlich tells the story, it became clear to experts in database management in the late 1990s that the exponential growth of data and increasing sophistication of software analytics and business intelligence were creating new demands on systems that were essentially designed for data warehousing. “As it turns out, database architectures that were designed for warehousing are terrible architectures for data analytics,” Ehrlich says. “Oracle can do lots of little questions, but stumbles when it comes to 10 terabytes of data and really difficult questions.”
The industry developed tools—Ehrlich calls them “crutches”— to make traditional databases work with software analytics. “With every new innovation,” Ehrlich says, “the demand [for analytics] went up and the innovations struggled”—with two notable exceptions. One was a decision at Teradata to develop data warehousing equipment based on massively parallel processing computer technology that ran Sybase IQ relational database software in what Ehrlich described as a “brute force” approach to the analytics problem. The other was Netezza, which was founded in
