Research On The Key Techonolgy Of Grid-Based Distributed Simulation System

With the rapid development of science and technology, it becomes more and morenecessary for researchers to use the techniques of modeling and simulation. The Modelingand simulation not only become an important tool in exploring the nature, discoverbiological mystery, opening aerospace field, and hovering the universe etc. but alsobecome an important method for people to share information resources and eliminate theisland of information. With the driving force from the sustained growth of the military, thebusiness and the requirement of scientific research, the modeling and simulationtechnology have experinced rapid development. Traditional distributed simulationtechnologies are not enough for people to share all kinds of simulation resources due tothe lack of load balance ability. Thus, the utilization rate of computational resource is lowand the system is hard to scale. Therefore, they can not adapt to the requirement of peoplefor large-scale complex simulations. With the expansion of the scale of simulation and theshart increase in requirements, these problems become the bottleneck in distributedsimulations.In order to overcome the shortcoming of traditional distributed simulationtechnologies, we propose the architecture of the high-level support system for gridservices. We divide the architecture into two parts: the grid-based RTI module and theinformation integration and management center module of RTI scheduling. Based on thearchitecture, we further propose the agent-based request callback mechanism, thegrid-based simulation module standard and the fault mechanism of grid-based simulationservice. The combination of the grid technology and the distributed simulation technologyprovides a solution for building the high-level support system of grid service. Therefore,the strategy is very practical in real applications.Existing simulation technologies usually passively choose tasks and service resourcesand lack the description of service characteristics of resources and the constraints ofservice ability. To cope with the problems, we propose an optimization mechanism forgrid-based distributed simulation tasks scheduling. In the mechanism, we first propose anoptimization scheduling theory model about service based on the characteristics of grid service. Then, we use the model to distribute tasks. Thus, it is impossible that the problemof performance bottleneck appears when the service resources are providing services fordifferent tasks. By using such mechanism, the problem about the unreasonable allocationof service resources and the poor scalability of system can be resolved.To address the problem of how to control the quality of service of system and managethe system resources, we propose a resource management mechanism mainly consideringthe quality of service. In the mechanism, we first set up a resource management modelbased on the resource management and its operation state. Based on the model, wepropose the forecast method about the quality of service of the system and the resourceallocation algorithm. All these constitute the management mechanism of system resourcesoriented the quality of service of system. By using the mechanism, the problem includingthe low quality of service and the lack of flexibility are resolved.In order to overcome the shortcoming of the current collection scheme of simulationdata, we propose a data resource mechanism according to the characteristics of grid-baseddistributed simulation. Specifically, in the mechanism, there is a data collection model, ananalysis method for data collection, a polymetization method for data collection, a storagemethod for data collection, a data collection module, a simulation application module ofclient, a data collection module of server, a data collection module of client, etc. In themechanism, the service is the main function. This makes the grid-based HLA distributionsimulation system have the good generality, interoperability and reusability.All in all, the system will have the features of high-level distribution, standards, highscalability and so on based on the above research results in the grid-based distributedsimulation system.