Research On The Real-time Fault-tolerant Issues In Distributed System Based On CORBA--Design And Implementation Of Dynamic Scheduling Scheme For Distributed System

Research on the distributed system has been carried out for a long time. To deal with the interoperability problem for the system, middleware was used. In the early-day's work, researchers focused on solving the problems of interoperability among heterogeneous platforms and environments, and seldom took account of attributes of real-time, dependability and others. In a word, the functions of distributed system were emphasized and its performance was ignored. With the development of the computer and the communication technology, the network bandwidth is expanded and the CPU speed is increased, and then people turn their attention to the performance of the distributed system which includes real-time, fault-tolerance, dynamic adaptability and combined real-time and fault-tolerance, etc.In the distributed system, if there are no measures supporting fault-tolerance, the real-time attribute of tasks can not be perfectly guaranteed. Therefore, how to design a novel architecture for real-time fault-tolerance system is becoming a hot topic for researchers, especially who deal with the mission-critical system. The trend is proved by the recent activity: OMG will issues the RT-FT CORBA RFP in the near future.In the paper, the solutions for the real-time fault tolerant dynamic distributed system are discussed in theory and in practice. The main works done by the author are listed below:The essentials of real-time and fault-tolerance in the distributed system are analyzed, and the necessity of their fusion is pointed out. Based on the work of other researchers, the idea used for combining the real-time and fault-tolerance in distributed system is proposed, and the solution for the problem is to use the RT-FT CORBA middleware.The real-time extension of middleware is the basis for the RT-FT middleware. Key technologies of RT CORBA middleware are discussed and a prototype of real-time middleware, rtORB, is developed. The latter's real-time performance is equivalent with TAO's reported by researcher abroad. For the dynamic attribute in the distributed system, the dynamic scheduling service is introduced into the real-time CORBA system. The two-tier end-to-end dynamic scheduling model is put forward, and its implementation and performance test are given out.The key technologies are discussed for the RT-FT CORBA middleware, including the three-tier RT-FT CORBA architecture, the extension of middleware, and the efficient dynamic scheduling service and algorithms to support the real-time fault tolerance.A prototype of RT-FT CORBA middleware is developed, which can be used for the real-time fault tolerance application, and also suit to the dynamic distributed environment.