Tape Group Parity Protection

We propose a new method of ensuring the redundant storage of information on tertiary storage, especially tape storage. Conventional methods for redundant data storage on tape include mirroring (storing the data twice) and Redundant Arrays of Inexpensive Tapes (RAIT). Mirroring incurs high storage costs, while RAIT is inflexible because special hardware is used and all tapes in a stripe must be loaded to read a file. Our method, Tape Group Parity Protection (TGPP), writes parity data asynchronously. The data area in each tape is divided into fixed-size regions. A collection of tape regions from different tapes are collected into a protection group. Parity data of the tape regions are computed, stored, and eventually migrated to tape. By using the disk cache to buffer the unmigrated parity data, all write operations can be performed asynchronously. Because each tape can be written independently, no special hardware is required to read the tapes, and individual tapes can be manipulated and transferred. In this paper we discuss TGPP, alternatives for managing TGPP groups, and their relative performance.