Energy-minimized scheduling for reliable workflow applications in heterogeneous cloud computing systems

With the rapid advancement of cloud computing, more and more modern large-scale workflow applications with different computing requirements are being migrated to clouds. Energy consumption has become a critical cost factor in computing systems, while reliability remains essential service quality metric. This research investigates the energy-efficient scheduling problem for reliable workflow applications in clouds. We propose a novel reliability distribution and energy-minimized algorithm (RDEMA) with heuristics for workflow scheduling. Firstly, a task clustering method based on task priority is introduced to simplify the complexity of task allocation. Secondly, a collaborative strategy based on different reliability constraint decomposition methods is proposed to meet the application reliability constraints. Finally, a heuristic energy reduction strategy which is based on dual-level dynamic voltage frequency scaling (DVFS) technique that operates at both processor and task granularities is developed for minimizing scheduling energy consumption. The proposed solutions are validated through extensive experiments using three real-world workflow applications and comprehensive synthetic parallel workflows, demonstrating significant energy savings while satisfying time and reliability requirements.