Ant Colony Optimization Approach for Distributed Online Scheduling

Online scheduling Algorithms play a challenging, complicated, and important role in the performance of distributed systems. This paper presents a decentralized adaptive online scheduling algorithm, named SfAntPM, with 3 main characteristics. First of all is SfAntPM's simple and novel pheromone definition and update rule, which is customized for online scheduling problems. The second is its Parameter Adaptation approach, with aim of contributing more diversification to search process. Combined with innovative probability model which is used to estimate the expected queue time, distributed scheduling agents used by SfAntPM can intelligently submit new arrival tasks to suitable resources independently. Empirical results show that SfAntPM outperforms other related algorithms, for performance metrics: makespan, load balancing factor as well as average weighted response time of tasks.