在异构系统中的基于并行编程模型的中间件设备.ppt

Middleware Infrastructure for Parallel andDistributed Programming Models inHeterogeneous Systems 1 INTRODUCTION The main objective: identify the common requirements for the parallel and distributed programming models and to propose and design a middleware infrastructure to satisfy these requirements. runtime support environment: intensive research in Java for parallel, distributed, and multithreaded programming models and applications on clusters and networked heterogeneous systems component：The distributed agents provide services: job distribution, monitoring, and controlling and so on for the system. Advantages: Portability, Expandability, Flexibility, Security and so on the rest of this paper: 1. CONCEPTS AND RELATED WORK 2. PARALLEL JAVA ON HETEROGENEOUS SYSTEMS 3. THE MIDDLEWARE INFRASTRUCTURE 4. FRAMEWORK FOR AGENTS’ SETUP AND ORGANIZATION 5. CONCLUSION AND FUTURE WORK CONCEPTS AND RELATED WORK Java’s popularity among developers is increasing steadily and many research groups are exploring the possibilities of using Java for high-performance parallel computing on multiprocessor systems. However, Many projects investigating parallel Java capabilities are in the research phase. Classify many projects into the four different categories Discipline: how they provide parallelism, their compatibility with the JVM, and user involvement。 1. Java dialects and preprocessors 2. Alternatives to JVM. 3. Parallelizing multithreaded applications 4. Pure Java implementations 2 PARALLEL JAVA ON HETEROGENEOUS SYSTEMS 2.1 Java Parallel Programming Models layer discipline： the level of user involvement in the parallelization process and the achievable levels of efficiency We first discuss the different parallel Java programming models 1. Message Passing the system provides some form of information exchange mechanism among distributed processes。 however