On a multi-objective evolutionary algorithm for optimizing end-to-end performance of scientific workflows in distributed environments

Large-scale distributed scientific workflows demand various system resources that are geographically scattered and typically shared by many users through wide-area network connections. These domain-specific workflows have different end-to-end performance requirements, which necessitate optimizing multiple objectives when mapped to heterogeneous network environments. We construct mathematical models for workflow mapping and formulate it as a multi-objective optimization problem to minimize latency and maximize throughput. We propose a workflow mapping solution based on a multi-objective genetic algorithm that uses a chromosome scheme to represent a set of possible workflow mapping schemes and employs genetic operators including selection, mutation and crossover to steer the evolution process. The performance superiority of the proposed mapping solution is illustrated through extensive simulations in comparison with existing workflow mapping methods.