Research And Improvement Of Digital Collaborative Computing And Simulation Platform For High-Speed Trains

Digital collaborative computing and simulation platform for high-speed trains (hereinafter called collaborative computing platform) provides a credible distributed computing platform for simulation of high-speed train. As the research continues and scale of simulation expanded, how to improve the maintainability and stability of the platform becomes an urgent task that needs to be solved.Through analysis of the problems that are encountered in deployment, test and use of the platform, System Management software are designed and new methods are proposed to improve performances of the Agent, Resources Management, Task Scheduler and Coupler subsystem, which led to enhancement of platform’ s deployment and management, improvement of resource utilization and system stability.System Management software is designed and realized based on the previous platform. It is used to manage the Agent subsystem, and through Agent subsystem it can control all these nodes which install it, shutdown or restart them; monitor the status of the Scheduler node; immigrate these nodes when they are failure; manage the program of scheduler, agent and third part applications, upgrade these programs when they have new versions; manage the processes of compute nodes; view the log files of these nodes and check work status and resource consumption of the task conditions that are submitted to the platform.Through analyse the problems of the previous Agent sub-system, the registration of the Agent is redesigned, which leds to the reduction of the resource costs and improvement of the registration reliability at the same time.’In order to monitor the status of the scheduler, maintain the information of the scheduler in database, close the Scheduler according to the order of the system manager, deploy the Scheduler software and upgrade the program, Scheduler agent program is designed. In this agent, program management function is used to respond the command of the System Management which upgrade agent software or install and start the third-party program; Log file management function is used to manage log files of each node; Node management function is used to respond System Management software’s command to shut down or restart computer; Process management function is used to obtain processes information of the computer and update it to database.For the problems of the scheduler subsystem, methods used to manage the resource and coupler nodes, alogorithm used to pre-schedule the task conditions in Scheduler have been redesigned, which led to the improvement of their efficiency in resource costs, task scheduling and accuracy of resource matching in advance. For the problems of the Coupler subsystem, iterative module is designed to deal with the conditions whose modules have iterative relationship with each other. The program interface is designed for Execute Machine when they want to communicate with the Coupler subsystem. Communication recovery algorithm is also designed for Execute Machine, which can greatly reduce the load of communication detection and recovery with the Coupler. So the platform’s stability and reliability is greatly improved.This paper firstly introduces the background and significance of the research, the current situation at home and abroad and the main research work related with the topic. Then it describes the function of each subsystem in the platform and flowchart of the submission for task conditions. The problems of the Agent, Resources Management, Task Scheduler and Coupler subsystem are analyzed particularly. Based on the analysis of these problems, improvement of three subsystems, design and realization method for the System Management software is described in great details. In the end, function tests are made for the whole platform. Furthermore, the whole work of the design is summarized and outlooked, the problems and next step research work for the platform are given at the same time.