UPS

The main storage issue at United Parcel Service (NYSE: UPS) used to be how to fit the most boxes in each truck. Now the company's storage focus is on maintaining critical data that allowed the company to ship and track 3.4 billion packages and documents last year.

"We evolved from a trucking company to a technology company," says Jim Medeiros, information systems and services manager for UPS, based in Atlanta. "Our business can't operate unless our data operates as designed, not only for uptime but for performance as well."

UPS calls what it does "synchronized commerce." It moves packages by truck and air to 200 countries and territories, allowing every customer who ships or receives a package via UPS to track its whereabouts every step of the way. The company also handles international payment through its own banking company.

The data that makes all that possible is stored in data centers in Atlanta and Mahwah, N.J. The data centers include 15 mainframes, 300 Tbytes of storage, and 1,500 midrange servers. In addition, UPS has 1,500 sort centers around the world that store another 400 Tbytes on 5,000 Intel-based servers. And that's not counting 105,000 tape cartridges on the network.

"We used to pick, pack, and ship," says Medeiros, who's in charge of UPS's data centers. "Now we're becoming a conduit to allow customers to leverage us, and our shippers want their customers to use the same data as we do. We invested $16 billion in technology over 15 years, and that provided us with a competitive advantage." [Ed. note: And, no doubt, tons of free vendor polo shirts!]Much of UPS's data center investment involved storage, which grew sixfold from 50 Tbytes in the early 1990s. Medeiros says the emphasis for storage is on reliability, cost management, and managing floor space.

UPS avoided the need for expanding the data centers to accommodate the growing storage needs in several ways, including switching to a SAN architecture two years ago. Now, UPS has 90 percent of its storage in a SAN, reducing the number of direct-attached storage (DAS) frames it operates from 165 to 130.

UPS used hardware and software from EMC Corp. and IBM Corp. for its SAN, which runs on Fibre Channel over Escon. "Our strategy is wherever possible to use two vendors," Medeiros says. "And EMC and IBM have software tools that manage their hardware."

In another move to reduce cost and floor space, UPS implemented a virtual tape solution from Storage Technology Corp. (StorageTek) over the past year. The network now has 105,000 cartridges on 24 StorageTek silos, saving 5,000 square feet in each data center and allowing Medeiros to reassign 20 technicians who were involved in managing disk storage.

Medeiros says DR/VFI (Disaster Recovery/Virtual File Identification) software from 21st Century Software also helps save money while providing business continuity. DR/VFI automates the disaster recovery process by determining critical files that must be backed up, and then carries out the backup."That had to be done manually in the past," Medeiros says. "It's brought a lot of surety to our disaster recovery process."

Medeiros says it would take 24 hours to copy data from one data center to another in case of disaster. UPS currently has a project underway to use Job Control Language (JCL), a command language for mainframe operating systems, to automate the backup process; Medeiros hopes it will reduce the backup time to four hours.

UPS has never been hit by a disaster that caused it to have to copy all of its critical data to the other center, "but you can't assume you'll never be in a situation where that data could be lost," Medeiros says.

Losing data would bring attention to the data center, which Medeiros wants to avoid. "Nobody cares about storage if it works," he says. "So I don't ever want people to think about storage."

That's his job.Dave Raffo, Senior Editor, Byte and Switch