An approach to cost-effective terabyte memory systems

The authors describe their approach for developing system architectures for secondary and tertiary storage systems and high-speed networks that will allow storage systems with 100-1000-Tbytes total capacity to become practical and widely used by the mid- to late-1990s. The research approach is founded on developing new techniques for managing latency, integrating compression, leveraging interleaving, and providing redundancy within the storage system. Some of the latency management strategies discussed include access hints and caching, reduced load times on tape media, and mixing optical disk and tape within the same hierarchy. The methods for increasing bandwidth include compression and interleaving. The authors test their ideas by building a prototype named Bigfoot that will store 10-100 Tbytes for Sequoia global change researchers. Bigfoot will provide a thousand-fold increase in online storage capacity over current disk-based systems, and it will provide a 10-100*improvement in capacity/cost over current tertiary storage systems.<>