Research On The Problems About Distributed Simulation Based On Grid Computing

In less than several decade, distributed simulation has gained a rapid development for its flexibility, efficiency and economical characteristics and has been applied in many fields such as avigation, military, social economics, entertainment and so on. The distributed simulation has been taken as a powerful tool in system analysis, design, testing, experiments, forecasting and operation training. Distributed simulation has become the third important method to understand and transform the objective world besides theory research and experiment. With a view to resolve interoperability and reuseability of distributed simulation, HLA has become the IEEE standard to construct distributed simulation. Distributed simulation based on HLA is applied more and more in military, avigation, weather, industry control, bioengineering and many other fields. With the extending use of simulation application, the scale and complexity of simulation system has increased, at the same time, the requirements for computing resources, storage resources, communication resources and special simulation instruments have also increased. Simulation application based on HLA is bothered more and more by some problems like relative low efficiency of computing resource usage, lack of dynamic load balance capability, lack of effective security mechanism, difficulty to save and deal with the vast data in time and effectively, which have become the bottleneck to construct the large scale distributed simulation tasks exactly and efficiently. In recent years, the grid computing and web services technologies provide a new approach to solve those problems of the traditional simulation based on HLA. The research of Grid-based HLA simulation technology has become a hot issue in simulation field. To resolve the above problems of HLA, by using grid computing technology this dissertation proposes a GADS framework to construct large scale and across-network distributed application, researches and discusses some key problems thoroughly. The main innovation achievements are listed as follow.By analyzing and researching the shortcomings of traditional distributed simulation framework, an advanced GADS framework is introduced, using the creative concept of simulation agent layer in order to improve the flexibility and capability of simulation under the GADS framework, the construction flow and the simulation interaction of distributed simulation had also been discussed.The simulation task scheduling service and load balance scheme of GADS had been studied and a schedule evaluation standard based on load balance and network traffic condition had been defined here. Then, taking it as a target function, simulation task scheduling algorithm based on UCG(Undirected Complete Graph) was proposed. Aimed at the characteristics that most distributed simulations are communications intensive, a communications preferred two stage scheduling algorithm based on UCG has been designed, and furthermore its advantage and application situation had been analyzed . As to dynamic scheduling problem,the definition of system load balance degree has been forwarded, and taking this as the basis, the load balance strategy under GADS framework was developed. By introducing simulation agent layer into GADS framework, the existing federate migratioin algorithm has been optimized and a freeze free federate migration protocol GFMP, which could just change the agent target of simulation agent in transfer process to ensure simulation source node and destination node overlap to some extent, has been designed. It would result in no need for suspending simulation execution during migrating, which would improve the efficiency of the whole simulation.As an optional service in distributed simulation system, HLA time management service provided many time management strategies, which are not mature enough. This dissertation researched thoroughly the present popular conservative time advanced mechanism, took the conservative time advanced algorithm based on lookahead and GALT as a example and analyzed its overhead and deadlock problem,then based on the analysis of classical Frederick algorithm, a improved conservative time advanced algorithm based on GALT has been designed which could avoid the deadlock appearing in Frederick algorithm. Aimed at optimistic time advanced mechanism, a checkpoint setting algorithm has been proposed which considered both rollback probability and rollback distance and through reasonable setup of checkpoints, the performance optimization of routine optimistic time advanced mechanism could be realized. Vector time was introduced creatively into the study of optimistic time advanced mechanism. Vector time updating scheme has been defined, moreover the conception of rollback vector has been introduced and those rollback algorithm based on vector time had been analyzed by using classical simulation advancing examples. On this basis, distributed simulation optimistic time advanced algorithm based on vector time has been designed. Considering the deficiency of HLA standard in simulation security management, distributed simulation security system structure was designed under the GADS framework and the security strategy under GADS frame was proposed. Based on trusted platform technology, we propose a group rekeying scheme which is stateless, needs no key server (i.e. distributed) and remains zero side effect on other members in the same group no matter which kind of change in membership happens. This algorithm could be applied not only in distributed simulation on GADS framework, but also in the other situations that need security group communication. In addition, a trusted transfer protocol based on trusted computimg for data transmission in open network environment has been designed. This protocol could be applied in federation migration across domain situation and also in other situations needing trusted transmission.On the basis of GADS simulation framework, a distributed testing simulation system was realized, which is one of the key experiment program of national defense in 2008 and has been put into preoperation. All performance parameters in this system have met the standard and the system is executed in satisfactory working order. This program is on the application for the first prize of Army Science and Technology Progress.