Based On Simulation The Solar Wind System Research The Massive Data Task Scheduling Algorithm

The solar wind is supersonic plasma charged particles ejected from the upperatmosphere of a star streams. The solar wind is a continuous existence, from the sunand the plasma200-800km/s velocity flow. High-speed solar wind particles formed bythe rapid changes in the flow, seriously affect the earth’s space environment, solarwind caused by geomagnetic storms, ionospheric storm will seriously interfere withthe wireless communication, the destruction of ground of large-scale infrastructureconstruction, various pipelines at the same time, the solar wind, will also have aprofound effect on the air space exploration. In order to avoid and reduce bring solarwind on the earth and the human disaster, should establish the physical andmathematical model for simulation of solar wind, solar wind system simulationexperiment, found that activity of the solar wind.Simulation of the solar wind system is not only a system of high computingperformance, but it is a massive data storage system. In the solar wind system in theprocess of simulation, simulation, the sun emits particles need simulation tasksthroughout the solar system, the simulation of the solar wind transport motionsimulation, the solar wind need simulation subtask of disturbances in the earth’smagnetic field simulation etc.. This makes the system simulation and performancedata processing task. In view of the characteristics of massive data task simulationsystem with the solar wind, requires the simulation platform is a distributedheterogeneous system platform, not only because of the large amount of datasimulation system, software and hardware system and distributed heterogeneoussystem has the characteristics of heterogeneous, will lead to the complexity of the taskscheduling strategy simulation system with massive data, and some tasks to the processor nodes it will have very high demand.The main goal of this subject is the massive data scheduling strategy simulationof solar wind system, because of better, massive data scheduling simulation system ofsolar wind capacity of computer processing faster, stronger requirements, distributedheterogeneous system will emerge as the times require, but given the large datacomplexity and the solar wind system simulation of distributed heterogeneous systemitself quantity characteristics and so, this also let scheduling in distributedheterogeneous system tasks become more complex.Research on scheduling algorithm, task model most uses the directed acyclicgraph DAG (Directed Acyclic Graph). Scheduling tasks, has been proved to be a NPcomplete problem. The most heterogeneous existing multiprocessor task schedulingalgorithm is firstly to initialize cluster, then the cluster is placed, and then to scheduletasks. Task scheduling algorithm based on task priority and copy algorithm based onthe clustering considering multiple factors of task execution time, the processor idlestate, precedence relations, causing the time complexity of the algorithm is too high.In view of this situation, research on earliest deadline scheduling strategy in thispaper, an improved algorithm based on the constraint of the precursor. In this paper, toinitialize cluster, proposed one kind based on the precedence constraint of the earliestdeadline cluster strategy to reduce the time complexity, in order to ensure thesuccessful scheduling rate, placed in the strategy and scheduling strategy forimprovement. The concrete work of the thesis are as follows:According to the same tasks on heterogeneous multiprocessor task scheduling inthe characteristics of different time in different processors, this paper proposed onekind based on the precedence constraint minimum execution time clustering strategy.This strategy will have a processor earliest deadline assignment of nodes to executethe same processor, only consider the node earliest deadline and precedence relations,in order to reduce the time complexity of clustering algorithm. When the existence ofindividual performance of the processor, based on the precedence constraint of theearliest deadline strategies will perform a large number of nodes to better processor.This situation will cause the individual processor overload, other processor idleproblem. To solve this problem, in place of each cluster to each processor tasks, usingload balancing placement strategy. This placement in the processor load balance at thesame time, can effectively improve the efficiency of task.In order to reduce the communication time delay, improve the utilization of processors, this paper improved the existing scheduling algorithms, by using theprecursor replication strategy based on reduce the communication delay, in order toimprove the task execution efficiency; and also used the redundancy removal strategy,delete the redundant nodes do not necessary, in order to improve the processorutilization.In order to evaluate the effectiveness of the algorithm in this paper, the ChengduUniversity of Technology solar wind system research laboratory’s massive solar winddata and the solar wind data processing system based on a lot of test analysisexperiment. Experimental analysis shows that the proposed precursor constraintsbased on earliest deadline scheduling strategy, the precedence constraint of the earliestdeadline clustering strategy effectively reduces the time complexity of the deletionstrategy based on, through multiple scheduling strategy and redundancy, in reducingthe time complexity of the algorithm at the same time, to ensure the schedulingsuccess rate, speedup.