Researching And Implementation Of The Intelligent Scheduling Strategy For Distributed And Parallel Computing

Multi-core computing, grid computing and cloud computing are a hot field of information technology in recent years and the core of these technologies is the scheduling problems of distributed and parallel computing. Scheduling model and scheduling algorithm are the most important aspects of task scheduling. Therefore, this paper presents task scheduling model and scheduling algorithm in distributed parallel computing environment. This major work is done by the following three aspects.1) Proposing a more practical task scheduling modelFrom the computing model, task model and performance model of the task scheduling, the existing task scheduling model is analyzed and summarized. Taking into account the model to be established to balance simplicity and accuracy, if the model is too precise, the models will be too complex which can not be resolved. If the model is too simple, then it can not be used for the actual system. This paper proposed and designed a more practical task scheduling model with the trade-off between practicality and accuracy.2) Designing three kinds of scheduling algorithms based on the practical modelBased-on the proposed more practical task scheduling model, this paper designed the three scheduling algorithms to solve the model. First, it is designed the simulated annealing scheduling algorithm and studied the performance of the simulation algorithm which is combined with the task scheduling model. Second, in order to find a more intelligent scheduling strategy, this paper proposed a self-adaptive genetic algorithm based on the practical model. The simulation shows that the algorithm has better performance and higher quality solutions. Finally, to simplify and improve the performance of scheduling algorithms, this paper studied how to use the ant colony algorithm to solve the problem based on the task scheduling model, the simulation shows that the algorithm has the best performance and the highest quality solutions.3) Implement a highly scalable distributed parallel computing platformAccording to the proposed task scheduling model, this paper adopts the ant colony algorithm which has better performance by comparing the simulated annealing algorithm, self-adaptive genetic algorithm and ant colony optimization. After determining the scheduling algorithm, this paper mainly thinks about the system implementation principles from the system functional requirements, the system architecture, the key detailed technologies of designing and system evaluation. The experimental results show that the system has strong versatility and high scalability for high-performance computing.