MySQL AB is partnering with Red Hat Inc. to ensure that the two companies support users running the MySQL open-source database on Linux as companies build out large server clusters and grids.

"We are making sure that we can support the horizontal scaling of systems" where users add low-cost servers to a cluster or grid of servers rather than buying bigger servers, MySQL CEO Marten Mickos said in an interview shortly after his keynote address to the annual MySQL user conference in Santa Clara, Calif., Wednesday. Mickos called the approach "scaling out, instead of scaling up."

At one time MySQL and Red Hat faced off as competitors. Three years ago, Red Hat launched a version of PostgreSQL as an open-source database under the Red Hat Database brand, but the effort flopped. Two years ago, it started distributing MySQL as part of its Red Hat Linux distribution and Wednesday the two formalized their joint support efforts.

Both want to cooperate with the proponents of grid computing, which tends to increase demand for both MySQL and Linux. ActiveGrid Inc., a 2-year-old San Francisco startup, launched software earlier this month for building scalable grids that run both Linux and MySQL.

Mickos cited Google Inc. as an example of a company that built its IT infrastructure around scaling out, instead of up. Cox Communications Inc. has built a data warehouse using 14 Linux server clusters to collect data and 25 Linux servers running MySQL to analyze the data, he said.

Sabre Holdings Corp., the owner of the Travelocity travel Web site, uses GoldenGate Software Inc. for data synchronization and management on a Linux and MySQL grid to dispense travel information, Mickos added. Friendster.com, an online social connections site, is based on a grid of 95 MySQL and Linux servers, Mickos said.

MySQL has found other partners, Mickos said, including Embarcadero Technologies Inc., a database tools supplier, and Emic Networks, a maker of high-availability and high-performance clustering software for open-source products, for building scaleable database clusters and grids.