SHF:Small: Crash Scene Investigation - Debugging Programs that Fail Unexpectedly

点击翻译按钮获取中文摘要

Computer systems are pervasive in all aspects of modern society. Thesoftware running on these systems, however, is difficult to implementcorrectly, and many failures occur after software is already released tothe public. Unfortunately, these "post-deployment" failures are oftenthe most difficult to fix. Failing runs are difficult to reproduce andanalyze, as information about these failures is much more difficult andcostly to obtain than failures occurring before software release. ThePI's research will address this problem by making post-deploymentfailures more informative without sacrificing the end-user experience.Specifically, the project will tackle this problem on two fronts. First,deployed programs will perform low-cost, customizable tracing to enhancecore memory dumps produced by failures. The project will investigate methodsusing both static and dynamic analysis to reduce the cost and improvethe utility of traced information. Second, the enhanced memory dumpswill require development of powerful new automated postmortem analysesto partially automate the process of debugging and programunderstanding. These analysis results can then be used to furtherinform useful tracing strategies via dynamic feedback from previousfailures. Thus, information about a particular post-deployment faultcould continually improve as more failures are observed. This automatediterative improvement of postmortem analyses has the potential tosubstantially reduce effort required to fix post-deployment faults, thusleading to faster and more accurate patches, resulting in greaterreliability for the software we all rely upon in our daily lives.