InformationWeek

Quantum Leap Into Action

The disk drive maker finds disaster rehearsals to be invaluable

By Bruce Caldwell
Issue date: May 15, 1995

The small Pennsylvania town of Wayne was enjoying unseasonably warm weather on Monday, Jan. 23. In spite of the spring-like conditions, six business-recovery workers from Quantum Corp. who had assembled at a nearby corporate park had blizzards and earthquakes on their minds.

At their feet on the raised floor of a hot-site data center were five plastic totes, sent by overnight express, packed with computer tapes. The team was waiting to load data onto H ewlett-Packard servers that had been configured to Quantum specifications.

Quantum, which is based in Milpitas, Calif., wasn't really preparing for a blizzard. But it was dead-serious about being prepared in case an earthquake or other disaster struck.

In addition, Quantum was convinced that its legal obligation to safeguard customers' records justified spending $300,000 annually for HP's disaster-recovery services.

In Case Of Emergency
In Wayne, the company was testing how well it could bring up 19 critical business systems, including financial, customer-service, and manufacturing applications. These systems run on two HP 3000 and 15 HP 9000 servers at its headquarters and 2,000 PCs and Apple Macintoshes at five sites worldwide.

Richard Blair, corporate data center manager for the $1.2 billion manufacturer, remembers when he realized just how valuable the disaster-recovery services can be.

Several days before the Pennsylvania test, Blair was in an airport waiting lounge i n Singapore. That was right after an earthquake had devastated Kobe, Japan. The Singapore airport was "crowded with survivors leaving and people trying to help," he recalls.

For Quantum, this year's rehearsal was more difficult than last year's. The company had acquired some disk-drive making businesses from Digital Equipment, but those operations run on VAX clusters at three sites. Each of the old Digital manufacturing sites has its own disaster-recovery plan. During the past year, Quantum had more than doubled the number of servers that needed to be recovered. These servers came online as the company migrated from systems running on the Ask Ingres v.7.0 relational database management and HP's proprietary MPE operating system to systems running on the Oracle 7.2 RDBMS and HP-UX.

HP, Quantum's disaster-recovery provider, had partnered with Weyerhaeuser Co. in Tacoma, Wash., in 1988 to offer disaster-recovery services using Weyerhaeuser facilities in Federal Way, Wash., and Wayne. HP provides the hardwar e, while Weyerhaeuser staffs the sites and offers services to clients with IBM platforms. Before this year's test, HP had configured the servers and tape drives at the Wayne hot site in preparation for the Quantum team's arrival.

HP sometimes runs rehearsals for customers, says Claude Brazell, U.S. program manager for business recovery services at HP in Santa Clara, Calif. But Brazell recommends that the customer come on site for the first two rehearsals to check plans and fine-tune disaster-recovery documentation.

Three Tasks
The Quantum team faced three tasks that Monday morning. Blair opened up the totes and began loading the tapes, one of the most time-consuming aspects of recovery. Network specialist Donald Whitely uncoiled cables and wires and began setting up ISDN connections to the Cisco routers. As software and networks started coming up, two database administrators and two systems managers checked configurations. Because Quantum's systems run on separate servers in Milp itas but on shared servers at the hot site, the group checked and dealt with naming convention conflicts.

After eight hours, 12 MPE applications were loaded, beating last year's recovery time by 10 hours. It would be Wednesday, however, before all seven HP-UX applications were loaded. According to on-site recovery experts, that time differential is common, because Unix recovery procedures and tools are still evolving. "There are always some 'gotchas' with Unix," says Darrell Dougall, a manager with Weyerhaeuser's recovery services in Wayne.

Quantum cut the loading time for its MPE applications by adding more tape units, memory, and disk space. Blair believes that by doing the same for the HP-UX applications-and adding the right mix of backup hardware and utilities-the HP-UX recovery time can be cut in half. He's also eyeing faster storage media. The digital audiotape he now uses records at 600 Kbps. Digital linear tape, a new format, could double that speed.

Blair will need the extra speed. By ye ar's end, some applications will be split on a new client-server system and Quantum will add five new servers. Says Blair: "We will have to define a disaster-recovery structure for distributed applications."

That means developing procedures and policies for remote sites so the tricky aspects of distributed computing can be managed more easily during a recovery. To prepare for its client-server disaster-recovery push, Quantum has scheduled a second rehearsal for August.

Comments on this story?