Distributed termination detection for dynamic systems

Abstract

A symmetric algorithm for detecting the termination of a distributed computation is presented. The algorithm does not require global information concerning the system and does not assume any communication features barring finite delays in the delivery of messages. It permits dynamic creation and destruction of processes participating in the computation. It provides for destruction of a process by external processes, such as the OS kernel. It also provides for external processes spontaneously joining an ongoing computation. Proofs of safety and liveness are provided.

Detection of Conjunctive Stable Predicates in Dynamic Systems

“…The critical points of correctness arguments are given as well. Although a number of elegant algorithms have been proposed and developed for stable predicate detection in static systems [1], [27]- [30] and several schemes have been proposed for termination detection in dynamic systems [10], [14], [20], [22], [36], our protocol detects general conjunctive stable predicates in dynamic systems. In addition, the system model assumptions employed in this study are less restrictive than those in previous works.…”

“…This may happen under certain situations. For example, a passive process may become active upon receiving a message in the case of termination detection [11], [13], [14], [20], [22]. In deadlock detection, a blocked process may be activated when a message granting a required resource is received [8], [41], [42].…”

“…On-line algorithms for monitoring global events using pair-wise interactions in peer-to-peer systems have been proposed by Chandra and Kshemkalyani [35]. Most of the existing schemes for detecting stable predicates are for static systems although several algorithms have been proposed for termination detection in dynamic systems [10], [12], [14], [20], [22], [36], [37].…”

Parallel SAT Solving on Peer-to-Peer Desktop Grids

“…More recent algorithms are applicable for completely asynchronous communication models where messages may arrive out-of-order or may be delayed for arbitrary but finite time [39][40][41][42]. While the most general of these algorithms support dynamic environments, many of them are not ready for use in real applications as they are restricted to scenarios where processors may be created but not destroyed [43][44][45] or require a node to participate in termination detection after it has been destroyed [46].…”

A general model for detecting distributed termination in dynamic systems

“…Since the works of Dijkstra and Scholten [11] and Francez [12], many elegant algorithms have been proposed to solve the problem in distributed systems. These include solutions for static systems [14,13,30,31,25,16,15,18,22] and dynamic systems [11,29,6,19,9,7]. Some algorithms take advantage of a global time base in the system [30,27,8].…”