In a replicated file system, copies of the same file are kept in different
servers sb that failures of some servers can be tolerant. The goal of a rep
licated file system is to increase the probability that an operation tread
or write) can be performed, i.e., to increase the availability of the file.
Giving a minimal acceptable availability of a file, the objective is to mi
nimize the total cost required to build a replicated file system that satis
fies the availability constraint. In this paper, an efficient algorithm is
proposed to find the minimal cost replicated file system. (C) 1999 Elsevier
Science Inc. All rights reserved.