USPS squeezes most out of tape

Megabyte for megabyte, tape may be the most affordable medium for storing electronic data, but its cost advantage can be diminished if data is not stored on the tape as efficiently as possible.

Megabyte for megabyte, tape may be the most affordable medium for storing electronic data, but its cost advantage can be diminished if data is not stored on the tape as efficiently as possible.

At the U.S. Postal Service Data Center in San Mateo, Calif., a virtual tape system is helping the center optimize its tape investment and create a powerful new way of backing up critical agency data.

One of two USPS data centers, the San Mateo facility hosts numerous applications designed to support operations such as product tracking and delivery logistics. As applications piloted at other USPS sites are approved, they are placed into operation for use at the center.

Beginning in 1998, the center began migrating mainframe data from its Storage Technology Corp. (StorageTek) tape platforms to new IBM Corp. Virtual Tape Subsystems (VTS), according to John Bodie, manager of software in the systems branch. The move was prompted by an "explosion in physical tape storage requirements," Bodie said.

"Moving to VTS allowed us to exploit the high density of the IBM Magstar tape media by consolidating several Storage-Tek tapes onto the Magstar tape," Bodie said. "That, in turn, let us save money in terms of provisioning floor space for tape drives and silos" instead of acquiring expensive new space.

VTS, one of a growing number of virtual tape products, uses a disk cache to receive and "pre-stage" data before it is written to tape. Pre-staging enables the full capacity of a tape to be used before another tape is written, thereby reducing the number of physical tapes that need to be stored in a library — and the number of libraries that need to be fielded.

Another advantage of VTS, according to Bodie, is speed. "We write data locally and direct a copy to the VTS cache," he said. That strategy achieves the speed of disk mirroring — which duplicates data from one disk to another — but delivers a copy of the data that can be stored off-site.

By the end of last year, the mainframe side of the data center had moved nearly 300 terabytes of data to VTS tape. Decommissioned StorageTek silos were handed down to another department at the center and redeployed.