About This Issue . . .

Nowadays the largest commercial database management systems are on the order of a terabyte. Emerging applications such as scientific databases for the Earth Observing System are expected to produce on the order of a terabyte of data per day. Applications requiring such data volumes also exist in the commercial arena, e.g., design databases. The objective is to achieve competitive efficiency in conjunction with a high-level query language interface, which provides efficiency of application implementation and maintenance. An issue of growing importance now is how to construct high performance, highly reliable systems out of a large number of components which individually have relatively small capacity and may be relatively unreliable. It is argued that the consensus problem is a basic building block in the construction of fault-tolerant systems. Applications of consensus algorithms to various problems in fault tolerant computing are given and the paper concludes with a vision of the role of the consensus problem in future systems design.