Recovery on the Cheap

Computerworld - The lights didn't go out in Oregon or Utah during last year's rolling blackouts, but officials at Portland, Ore.-based PacifiCorp were worried anyway. If an outage were to hit the company, it would have no way to recover the mission-critical decision-support system that schedulers and traders rely on to buy, sell and schedule power loads at the $4.2 billion electricity provider. The best way to truly protect the trading system was via real-time data replication of the company's complex Oracle database. But in a tight utility market, there was just no way PacifiCorp could afford a separate backup system.
"Our priority was to keep costs minimal . . . so we didn't even consider buying a separate platform for disaster recovery," says Steve White, manager of Unix and database administration at PacifiCorp.
Instead, White and his team opted to split the power-scheduling system between the company's two existing data centers: a production facility in Portland and a testing facility in Salt Lake City. Now the scheduling system runs primarily in Portland, with real-time data replication to Salt Lake City.
It's a data recovery solution that the power-trading unit is happy to pay for - $75,000 to $100,000 for software and storage, vs. the millions it would have cost for a stand-alone backup system that would sit idle most of the time.
PacifiCorp isn't alone in trying to squeeze more processing power from existing resources and new technologies to provide disaster recovery capabilities on the cheap.
Instead of housing all IT at a single site with a second site or an outside service provider lined up for backup, a growing number of companies of all sizes are splitting IT between two locations, says Cal Braunstein, an analyst and director of research at Robert Frances Group Inc. in Westport, Conn. For example, instead of having 100 servers at one site, a company might opt for two locations and place 75 servers at each, which involves an additional expenditure for the extra servers but could result in a cost-effective backup site. Although in most cases the second site would primarily be used for testing, it could run enough of the business processing to operate during a disaster, explains Braunstein.
Taking a page from the playbook of many of its customers, Veritas Software Corp., which sells storage management software, is also splitting data processing by using new Fibre Channel data-transmission capabilities to connect its Mountain View, Calif., headquarters and a Minnesota engineering center. "With more recent technologies, you can move data over any distance,