Multi-Objective Genetic Algorithm for Tasks Allocation in Cloud Computing
The problem of allocating real-time tasks to cloud computing resources minimizing the makespan is defined as a NP-hard problem. This work studies the same problem with two realistic multi-objective criteria; the makespan and the total cost of execution and communication between tasks. A mathematical model including objective functions and constraints is proposed. In addition, a theoretical lower bound for the makespan which served later as a baseline to benchmark the experimental results is theoretically determined and proven. To solve the studied problem, a multi-objective genetic algorithm is proposed in which new crossover and mutation operators are proposed. Pareto-optimal solutions are retrieved using the genetic algorithm. The experimental results show that genetic algorithm provides efficient solutions in term of makespan for different-size problem instances with reference to the lower bound. Moreover, the proposed genetic algorithm produces many Pareto optimal solutions that dominate the solution given by greedy algorithm for both criteria.