分布式调试中基于事件模型的确定性重演策略  

Strategy of deterministic replay debugging based on the event model in distributed debugging

作　　者：李青山[1] 李珺[1] 叶宏[2] 杜林[1] 

机构地区：[1]西安电子科技大学软件工程研究所,陕西西安710071 [2]西安航空计算技术研究所,陕西西安710068

出　　处：《西安电子科技大学学报》2010年第5期872-878,共7页Journal of Xidian University

基　　金：航空科学基金资助项目(20081981010;20080531005);国家"十一五"国家部委预研资助项目(513060401)

摘　　要：分布式系统中,进程并发执行的不确定性导致了程序错误的不可再现性,后续的执行无法再现前次执行的错误,这使得反复执行程序以重现故障的循环调试方法不再可用.基于事件模型的记录重演机制,提出了一种确定性的重演策略.该策略使得程序的执行轨迹确定化,重现了程序初始运行的错误状态,并结合传统的循环调试策略,通过设立断点或单步调试等方法最终定位错误.In the distributed system,the nondeterministic feature that the process＇s concurrency causes the error can not be reproduced.The following execution of the program is unable to reproduce the mistakes that occurred on the previous execution.This causes the circulation debugging method no longer to be available.Based on the event model,by studying the mechanism of record and replay during the debugging process,a strategy for deterministic replay is given.This strategy can make the path of the procedure deterministic,reproducing the program initial error status.Then through the traditional circulation debugging strategy,it is possible to set the breakpoint or to debugging step by step to locate the error.

关 键 词：分布式调试 事件模型 不确定性 重演 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]