ALCOMFT-TR-03-115

Kay Salzwedel
Algorithmic Approaches for Storage Networks
Paderborn. Work package 1. December 2003.

Abstract: The last level of the storage hierarchy in modern computers consists of reliable, persistent storage devices. Naturally, this level has the slowest access time, the largest capacity, and the cheapest costs. Nowadays, mainly magnetic hard disks are used to store large amounts of data permanently. But even the constantly growing capacity of disk drives cannot satisfy the ever growing demand for flexible and efficient storage systems. The simplest way to provide almost unlimited storage capacity is the use of disk drives in parallel. The underlying architecture of such a storage network may vary significantly. All it requires is a number of storage devices connected by an arbitrary network. Naturally, this network also provides communication links to the outside world.

Unfortunately, the efficient use of storage networks is not straightforward. There are a number of requirements, like data distribution, balance of data accesses, heterogeneity, or adaptivity, which have to be met to exploit the full potential of storage networks. In this article we give an overview of existing techniques and algorithms addressing these problems. We start with classical solutions, like striping techniques and the use of redundancy, but the main focus of this work will be on more challenging tasks - namely heterogeneity and adaptivity. There are a number of solutions for each of these problems, but the most daring situations arise when all requirements must be met. We will introduce some techniques that can handle many of these constellations efficiently.

Postscript file: ALCOMFT-TR-03-115.ps.gz (153 kb).