This paper investigates the problem of error propagation in distributed sys
tems. To resolve this problem, a state preservation scheme is presented to
save process states in main memory. Based on the state preservation, the pr
ocesses suffering from error propagation can be recovered without involving
stable storage. The recovery overhead is significantly reduced. In additio
n, a well-known single-source-all-destination graph algorithm is also utili
zed to find the optimal recovery points of the processes suffering from err
or propagation. (C) 2000 Elsevier Science B.V. All rights reserved.