Abstract: A computing resource sharing and aggregation platform in the heterogeneous environments called DCSA is presented and described. DCSA is a parallelization platform that enables the programmer to build distributed parallel applications using a high level and easy-to-use parallel programming paradigm. DCSA is designed for a general system by defining XML approach with which the application developer can document and evaluate the results of program. A light-weight check point mechanism and a highly customizable communication mechanism are presented that prove the fault tolerance and efficiency of DCSA. Results obtained from examples analysis show that DCSA is feasible and efficient computing resource sharing platform.

Key words: high performance computing, resource sharing, distributed computing, multi-thread

摘要： 描述异构环境下的空闲计算资源共享与聚集平台DCSA。DCSA是一种通用的、可扩展与高效并行化的平台，在该环境下可以较容易地使用高级别的方式编写并行程序，把各种不同的软硬环境下的机器整合到单一的系统环境，共同完成高性能科学计算问题。DCSA的通用性得益于统一的XML文档来描述任务与计算结果及灵活的编程模型。利用一个轻量级的检查点机制与高度可配置的通信机制保证了DCSA的容错性与高效性。通过对实际并行例程的案例分析，验证了DCSA的正确性与高效性。

关键词: 高性能计算, 资源共享, 分布式计算, 多线程

CLC Number: 

• TP338.8