Files in this item



application/pdfJin_Dongyun.pdf (2MB)
(no description provided)PDF


Title:Making runtime monitoring of parametric properties practical
Author(s):Jin, Dongyun
Director of Research:Rosu, Grigore
Doctoral Committee Chair(s):Rosu, Grigore
Doctoral Committee Member(s):Havelund, Klaus; Agha, Gul A.; Marinov, Darko
Department / Program:Computer Science
Discipline:Computer Science
Degree Granting Institution:University of Illinois at Urbana-Champaign
Subject(s):Runtime Monitoring
Parametric Monitoring
Monitoring Oriented Programming
Abstract:Software reliability has become more important than ever in recent years, as a wide spectrum of software solutions are being used on various platforms. To this end, runtime monitoring is one of the most promising and feasible solutions for enhancing software reliability. In particular, runtime monitoring of parametric properties (parametric monitoring) has been receiving growing attention for its suitability in object-oriented systems. Despite many parametric monitoring approaches that have been proposed recently, they are still not widely used in real applications, implying that parametric monitoring is not sufficiently practical yet. In this dissertation, three perspectives for better practicality of parametric monitoring are proposed: expressiveness, efficiency, and scalability. A number of techniques on all three perspectives are developed and integrated to the JavaMOP framework, which is a formalism-independent, extensible runtime monitoring framework for parametric properties. One limitation in expressing parametric properties is that the first event must alway initiate all parameters. This limitation is removed in the proposed work to improve expressiveness of parametric monitoring. Further, a new logical formalism, PTCaRet, is introduced for describing properties of the call stack. As for efficiency, the `enable set optimization', the `indexing cache', and the `monitor garbage collection' are proposed for optimizing creation of monitors, access to monitors, and termination of monitors, respectively. In addition, several scalable parametric monitoring techniques are introduced. These techniques, for the first time, allow a large number of simultaneous parametric specifications to be monitored efficiently. The optimization techniques presented in this dissertation were implemented into the JavaMOP framework, yielding JavaMOP 3.0, the latest and most efficient version of JavaMOP. Thorough evaluations show that these techniques can improve runtime performance of JavaMOP by 3 times on average, and up to 63 times in some cases; as for memory usage, by 3 times on average. While Tracematches and the previous version of JavaMOP crashed on several cases due to out of memory errors, the newer version of JavaMOP did not crash on any case during the evaluations. Considering that the previous version of JavaMOP was one of the most efficient parametric monitoring frameworks in terms of runtime performance, the results presented in the dissertation can be argued significant.
Issue Date:2012-09-18
Rights Information:Copyright 2012 Dongyun Jin
Date Available in IDEALS:2012-09-18
Date Deposited:2012-08

This item appears in the following Collection(s)

Item Statistics