Locating Features in Distributed Systems

In distributed systems, just as in conventional software, it is often necessary to locate the software components that implement a particular user feature. Several dynamic analysis methods have been proposed to address this feature location problem in conventional software. Most compare traces of execution that exercise different combinations of features. The feature location problem for distributed systems has complexities beyond those found in sequential systems; namely, concurrent processes and lack of a total ordering of events. This paper introduces a dynamic analysis technique for distributed systems that addresses these complexities. Our methodology defines a component relevance index that can be computed for each software component in the system. Repeated execution of the feature yields more precise indexes for the components. The software components can then be ranked to identify those most relevant to the task at hand. A small case study is presented to illustrate the formal...

• Type: Conference
• Year: 2004
• Where: SERP
• Authors: Sharon Simmons, Dennis Edwards, Norman Wilde
• Keywords: Distributed Systems, Feature Location Problem, SERP 2004, SERP 2007, Software Components