Maximizing Reliability of Heterogeneous Distributed System Using an Adapted Discrete Flower Pollination Algorithm for Task Allocation Problem

Finding task allocation that maximizes reliability of a heterogeneous distributed system is an NP-hard problem. For that, meta-heuristic is used to get a sub optimal solution in reasonable time. The Flower Pollination algorithm is new meta-heuristic used successfully for solving different problems in different fields. The original Flower Pollination algorithm is designed to deal with continuous problem. Thus, applying this algorithm to discrete problems in its original form seems to be very hard or useless. In this paper, we propose an adapted discrete version of the Flower Pollination algorithm to deal with the problem of maximizing reliability of a heterogeneous distributed system under task allocation problem. This algorithm is called Adapted Discrete Flower Pollination algorithm (ADFP). To confirm the effectiveness of our algorithm, we have tested and compared its results with that of Hybrid Particle Swarm Optimization (HPSO). The Experiments results show the effectiveness and superiority of ADFP over HPSO in all tested cases.