M-Guard: A new distributed deadlock detection algorithm based on mobile agent technology

Abstract

Deadlock detection and resolution are of the fundamental issues in distributed systems. Although many algorithms have been proposed, these message passing based traditional solutions can hardly meet the challenges of the prevailing Internet computing and mobile computing. In this paper, we present a novel algorithm, namely the M-Guard, for deadlock detection and resolution in distributed systems based on mobile agent technology. The proposed algorithm lies in the intersection of the centralized type algorithm and the distributed type algorithm. An agent is employed in our algorithm as a guard with dual-role: when roaming in the system according to a specified itinerary algorithm, the agent collects resource request/allocation information for detecting deadlock cycles as well as propagating the collected network and resource information among the nodes. Consequently, accurate and timely detections of deadlocks can be made without any network node being the performance bottleneck. Preliminary simulation results show that, compared with several other algorithms, the M-Guard algorithm achieves both shorter deadlock persisting time and smaller phantom deadlock ratio. Moreover, the overall network communication overhead can be decreased, too. © Springer-Verlag Berlin Heidelberg 2004.