Slicing Event Traces of Large Software Systems

Abstract: Debugging of large software systems consisting of many processes accessing shared resources is a very difficult task. Many commercial systems record essential events during system execution for post-mortem analysis. However, the event traces of large and long-running systems can be quite voluminous. Analysis of such event traces to identify sources of incorrect behavior can be very tedious, error-prone, and inefficient. In this paper, we propose a novel technique of slicing event traces as a means of reducing the number of events for analysis. This technique identifies events that may have influenced observed incorrect system behavior. In order to recognize influencing events several types of dependencies between events are identified. These dependencies are determined automatically from an event trace. In order to improve the precision of slicing we propose to use additional dependencies, referred to as cause-effect dependencies, which can further reduce the size of sliced event traces. Our initial experience has shown that this slicing technique can significantly reduce the size of event traces for analysis.