Research On Energy-saving Parallel Application Scheduling Optimization Algorithm In Heterogeneous Distributed Systems

With the rapid expansion of computer scale and the continuous improvement of performance,the energy consumption of computer systems is also increasing rapidly.For heterogeneous distributed systems,energy consumption has always been one of the main considerations for large cluster computer systems or small handheld terminal devices.With the rapid increase in energy consumption for complex performance,many efforts and work are focused on ensuring scheduling timeliness while taking into account the optimization of the system energy consumption.How to balance the energy consumption and performance has become a major challenge in academia.For the task scheduling problem,how the application can make full use of the processor in the computer system for high performance computing has become a practicle research field.The main purpose of the existing parallel task scheduling optimization mechanism is to reduce the scheduling length of the entire parallel application,while improving the operating efficiency of the system and balancing the system load to a certain extent.The basic idea is as follows: firstly,the application is divided into multiple subtasks,and then based on a certain scheduling strategy,these subtasks are assigned to the appropriate processor to get the fastest response.At present,Dynamic Power Management(DPM)and Dynamic Voltage/Frequency Scaling(DVS/DVFS)are the most widely used in task scheduling.The DPM technology is mainly proposed to save the energy consumption when the system is running,and this technology achieves the goal by setting the currently long-time idle components of the system to a low power state.The idea of energy consumption optimization in DVS/DVFS technique is to reduce the energy consumption of the processor by changing the operating voltage and clock frequency of the processor during the execution of the system.Based on the existing algorithms,this thesis proposes an enhanced energy consumption constrained parallel application scheduling algorithm(EECC)in a heterogeneous distributed environment.The main idea of the algorithm is divided into two parts.First,it satisfies the energy consumption constraint given by the parallel application,which is solved by transferring the energy consumption constraint of the application to each task in a certain way.Second,reduce the scheduling length by heuristic scheduling each task with low time complexity to solve.In order to verify the effectiveness of the algorithm and ensure the accuracy of the experimental results,this thesis uses Java language to implement a simulator,simulates the parallel applications in real heterogeneous distributed environment through three different models,and evaluates the EECC algorithm from different angles.The experimental results show that the proposed algorithm can achieve better scheduling length while satisfying the energy consumption constraints,and has better performance than the existing parallel algorithms.