Achieving mutual exclusion is one of the most fundamental problems in distr
ibuted computing. The use of coteries is a well-known approach to this prob
lem. A coterie is a special set of pair-wise intersecting node groups calle
d quorums. The communication delay incurred in a quorum-based mutual exclus
ion scheme depends critically on the coterie adopted, and thus it is import
ant to find a coterie with small delay. Recently, two related measures call
ed max-delay and mean-delay have been introduced. The former measure repres
ents the largest delay among all nodes, while the latter is the arithmetic
mean of the delays. In a previous paper, we have proposed a polynomial-time
algorithm to find max-delay optimal coteries, but there has been no algori
thm to find mean-delay optimal coteries. In this paper, the first algorithm
that finds mean-delay optimal coteries in general topology networks is pro
posed. This algorithm employs the branch-and-bound method to effectively ex
plore the search space. Although its running time can be exponential, it is
shown that the algorithm is applicable to moderate-sized networks through
experiments. (C) 2001 Elsevier Science Inc. All rights reserved.