摘要： 当前并发程序容错机制处理方式单一、效率较低。为此，提出一种适用于多种并发程序错误处理的容错机制。通过在编译及运行过程中对程序进行异常处理，并在异常发生时根据设置的检查点对程序进行回滚和防错误处理，以实现并发程序容错。实验结果表明，该容错机制可有效检测并发程序中的错误，在不增加程序总体运行时间的情况下达到比较理想的容错效果。

关键词: 异常处理, 容错, 并发程序, 多线程, 原子性违例, 回滚, 检查点

Abstract: Currently most concurrent program fault-tolerant mechanisms can not handle multi-bugs in programs and are also of low effectiveness. This paper presents a concurrent program fault-tolerant framework based on exception handling, which can be used to handle different concurrent program bugs by supporting from compile-time and runtime work and rollbacking program status to certain checkpoints in order to achieve the objective of fault-tolerance of concurrency bugs. Experimental results show that this framework can help to detect concurrency bugs effectively and recover program states from exceptions to normal states.

Key words: exception handling, fault-tolerance, concurrent program, multi-threaded, atomicity violation, rollback, checkpoint

中图分类号: 

• TP391