Research On Simulation Computing Task Scheduling Method For Numerical Tank

The Numerical Tank Simulation Platform is a system for numerical simulation of hydrodynamics in ship and marine engineering.It integrates various simulation subsystems based on high performance computer clusters and applies CFD technology to complete the physical tank requirements simulation,to provide users with virtual simulation experiment application technology.In order to improve the task execution efficiency of numerical tank platform and improve the utilization of platform resources,this thesis studies two aspects: one is the intelligent allocation between tasks and virtual machines;the other is the resource scheduling method for balancing the load of physical machine cluster.Based on the analysis of task characteristics of numerical tank platform and the characteristics of resource requirements,this thesis proposes a method to evaluate the task size to get the task priority queue.Due to the long running period of platform tasks,the state of a system at a moment in time can't represent the load of the current system and the load of the virtual machine needs to be evaluated.In response to this problem,this thesis presents a dynamic load evaluation model to get the load of virtual machines.Based on this model,a task scheduling algorithm based on median and least load is proposed to match the task with the virtual machine.The performance of the algorithm is verified through experiments.In order to improve the resource utilization of the numerical tank platform,the problem of unbalanced load on the physical cluster is solved by virtual machine migration,and an empty physical machine is shut down to save energy by the method.In this thesis,an improved Particle Swarm Optimization(PSO)algorithm is proposed to migrate virtual machines.This algorithm considers the resource utilization of the entire physical cluster and the consumption of the virtual machine migration,and finds the correspondence between the virtual machine and the host physical machine to be migrated,so that the resource utilization rate is the highest and the migration cost is minimum.The performance of this algorithm is verified by experiments.