Deadlock avoidance: Improved algorithms for centralized and distributed systems

Abstract

A deadlock avoidance algorithm for a centralized resource allocation system is presented. Unlike the Banker's algorithm, this proposed algorithm makes use of the state of the previous safe sequence to construct a new safe sequence. The performance of this proposed algorithm is compared to that of both the Banker's algorithm and an efficient algorithm proposed by Belik. The simulation results show that our algorithm's execution time is significantly better than the Banker's algorithm and is very competitive with Belik's algorithm. In addition, our Modified Banker's Algorithm produces optimal results unlike Belik's approach which sometimes deems a safe allocation request unsafe. This centralized algorithm combined with an algorithm by Moser, is extended for use in distributed systems. Compared with Moser's algorithm, this algorithm is less restrictive in acquiring resources from other processes and allowing increase in its maximum resource requirement as confirmed by our analysis and simulation results.