Detecting the termination of asynchronous distributed cooperative computations

This article presents an algorithm for detecting the termination of asynchronous distributed computations in cooperative environments. It is similar to the general parallel termination problem, which has been thoroughly studied in the literature. However, the problem discussed in this article has a different nature and setup. One major difference lies in the fact that in most cases the possible communications between processors (computers) would result in an unpredictable forest pattern. Another difference is our assumption that all processors in the system operate in an cooperative environment to finish one large task. In this sense, no processor (meaning a respective process or thread running on it) is allowed to abandon the operation and start a new one until all have terminated. Initially, this problem was encountered during the design stage of a parallel database synchronous rule-processing system. Hence, the motivation for the solution proposed in this article originates from a practical necessity. Nevertheless, such a problem may be encountered in many evolving cooperative computational applications that utilize multiple computers over the Internet to perform large computational tasks.