Research On High Performance Computing Application Model For Spacecraft Simulation Analysis

The input of resource cost of simulation calculation becomes the most important part of enterprise cost input,which is second only to the cost input of human resources.Although the high-performance computing equipment is purchased,the full use of equipment can’t be guaranteed,making the enterprise competitive ability be obviously insufficient.Based on the project background of the spacecraft manufacturing enterprise,combined the resource dispatching system LSF,that how to minimize the time span of task set while considering computational resource load balance are studied in the paper which start from constructing high-performance computing environment,optimizing job scheduling algorithm and enhancing system resource utilization.The selection of job scheduling algorithm is the key of job scheduling.As a classic heuristic scheduling algorithm,Min-Min algorithm is often used as a standard algorithm to evaluate other algorithms.In this paper,the further research is carried out based on Min-Min algorithm.First of all,the Min-Min algorithm is studied and analyzed in detail.Its basic idea is to give priority to the computing task with the smallest earliest completion time,thus realizing the shortest execution time of a single task,which essentially is a greedy algorithm.However,due to the defect of greedy algorithm,the current choice may be the local optimal solution.In order to minimize the minimum completion time,the algorithm sometimes needs to allocate the task to a computing node with a long execution time,resulting in an increase in the total time span.At the same time,the algorithm first maps the task to the computational resources with the strongest computing power,resulting in heavy computing resource load,weak computational resources idle and unequal computing resource load,which leads to low resource utilization.Secondly,in view of the above shortcomings,in this paper,the Min-Min algorithm is improved,and a minimum earliest finish time algorithm(SDMM)based on standard deviation is proposed.In the manufacturing industry,the cluster system is often constructed in stages,so the computing environment has a certain difference and isomerism,and the task is different in the execution time of the different processors.In this paper,task execution time standard deviation is used to measure the difference between different processors and the influence of tasks on load balancing.The algorithm is scheduled on the basis of the minimum completion time and the standard deviation of the task execution time.It takes both the high throughput rate and the load balance,and is more adaptable to the heterogeneity of the production environment.Finally,in this paper,the Min-Min algorithm,the Dpri-Min-Min algorithm,the Proposed algorithm and the SDMM algorithm are simulated and studied through the GridSim simulator,and the result shows that SDMM algorithm has some improvement compared with Min-Min algorithm,Proposed algorithm and Dpri-Min-Min algorithm.The high performance of the original algorithm is preserved,the load balance is realized,the license resource utilization is improved,and the actual demand for cost control and resource lean utilization in the simulation of spacecraft is met.Therefore,the research on high performance application model and scheduling algorithm for spacecraft simulation analysis has a certain reference significance for large manufacturing enterprises such as aviation and aerospace.