Multi-Installment Scheduling Optimization Model And Algorithm With Result Collection

Research on task scheduling strategy has always been a hot topic in distributed computing platforms.The goal of distributed platform task scheduling is to find a reasonable task scheduling stratrgy to make the task makespan the shortest,and the task makespan includes transmission time,computation time and result collection.Most studies have assumed that the collection time is negligible,however many large data applications have completed task computation on distributed platforms,and the data size of their results is still large(such as image compression),that is,the transmission time of the results cannot be ignored.In view of this,taking the shortest makespan of tasks as the goal,this paper studies different scheduling models with results collection under two communication modes,and designs efficient algorithms to solve the models.The main contribution of this paper can be summarized as follows:1.Aiming at the heterogeneously distributed platforms,we investigate the scheduling problem of multi-installments with result collection under blocking communication mode.Firstly,in order to minimize task makespan,a new multi-installments scheduling optimization model is established.Secondly,under the given server scheduling sequence,the optimal scheduling number and the optimal task assignment scheme are derived.Then,a new global optimization genetic algorithm is designed to obtain the optimal scheduling sequence.Finally,by comparing with the existing multi-installments scheduling algorithms,the proposed algorithm can achieve the shortest task makespan,and the validity of the model and algorithm is verified.Moreover,the experimental results show that when the task size is large,the optimal scheduling sequence of the server is the ascending order of the server transmission rate.2.Aiming at the heterogeneously distributed platforms,we investigate the scheduling problem of multi-installments with result collection under non-blocking communication mode.Firstly,a new multi-installments scheduling optimization model is established to minimize the task makespan.Then,based on the conclusions of the above chapter,the scheduling sequence of the server is fixed as the sequence of increasing transmission rate,and the optimal scheduling number and task assignment scheme of multi-installments scheduling in non-blocking communication mode are derived.Finally,the validity of the proposed model and algorithm is verified by experiments.Experimental results show that the proposed model can not only minimize task makespan,but also greatly improve the fault tolerance performance of distributed platform.