The surest way to beat the competition is to start fast and stay ahead. That strategy made Taleo a 2006 Enterprise All-Star Award winner for the virtualised application infrastructure it completed in June 2005.
IT executives at the San Francisco human resources firm bet on Linux (ditching Unix) when they were planning the new system in 2002. "At the time, Linux was being looked at as a way to save money at the Web tier, not something that you would run mission-critical applications on across all tiers," recalls Brad Benson, CTO at Taleo.
IT executives were challenged to find ways to increase business while keeping IT costs low and performance high. Taleo offers on-demand, hosted applications for a range of job-recruiting needs. Of its roughly 500 employees, nearly half are in IT-related roles.
By 2004, Taleo had rolled out Linux on its Web and application servers and in 2005 tackled the difficult database layer. The database project focused on PolyServe's Database Utility for Oracle, which clusters databases on low-cost x86 servers running Linux and a storage-area network (SAN). "PolyServe helped us solve device-driver issues and other roadblocks," Benson says.
The team deployed multiple database clusters with four to six servers per cluster. And, it standardised on HP Proliant blade servers for the application layer - more new technology of the day - tapping into its Fibre Channel SAN. The result was a state-of-the art application infrastructure that can scale as use demands, without a hiccup in performance.
"In this shared data arrangement, the services of any server can automatically transition to any other server in the cluster with no loss of data or performance," Benson says. "We can scale database server performance and capacity by adding another HP server while the cluster is online."
Other organisations only now are deploying enterprise-scale Linux and tinkering with application virtualisation in the lab, but Taleo has been reaping the rewards of its virtualised databases for more than a year. The environment has boosted hardware efficiency, with each server now handling 25 per cent more work, and it has improved reliability by one nine to 99.9 per cent, Benson says.