Algorithmic Approaches for Storage Networks

Abstract

Persistent storage in modern computers is usually realized by magnetic hard disk drives. They form the last, and therefore the slowest level in the memory hierarchy. Disk drive technology is very sophisticated and complex, making an accelerated growth in disk capacity possible. Nowadays, a single of-the-self disk drive is capable of storing up to 180 GB and this number is doubled every 14 – 18 month. Nevertheless, this is not sufficient to manage the ever growing data volumes. The fast evolving processing power of modern computers, the global availability of information, and the increasing use of mixed-media contents demand for more flexible storage systems. The easiest way of providing almost unlimited storage capacity is the use of many disk drives in parallel. Unfortunately, a straightforward solution does not exploit the full potential of storage networks. As an example, suppose there is the need to add further disk drives to an existing system due to changed capacity demands. Without a redistribution of already stored data the capacity problem is solved but the addition of new (and usually faster) disks does not improve the overall performance of data accesses. There might exist very popular data on an ’old’; disk resulting in a heavy load for that disk and leading to a bottleneck in the system.

Partially supported by the Future and Emerging Technologies programme of the EU under contract number IST-1999-14186 (ALCOM-FT).