Some users are experiencing upload errors at the moment. If you receive a "UWSpace is down for maintenance" error, please email jordan.hale@uwaterloo.ca as soon as possible. We are very sorry for the inconvenience.

Show simple item record

dc.contributor.authorWu, Chun Wah Wallace
dc.date.accessioned2013-01-18 16:59:02 (GMT)
dc.date.available2013-01-18 16:59:02 (GMT)
dc.date.issued2013-01-18T16:59:02Z
dc.date.submitted2013
dc.identifier.urihttp://hdl.handle.net/10012/7215
dc.description.abstractRuntime verification is a lightweight technique that serves to complement existing approaches, such as formal methods and testing, to ensure system correctness. In runtime verification, monitors are synthesized to check a system at run time against a set of properties the system is expected to satisfy. Runtime verification may be used to determine software faults before and after system deployment. The monitor(s) can be synthesized to notify, steer and/or perform system recovery from detected software faults at run time. The research and proposed methods presented in this thesis aim to reduce the monitoring overhead of runtime verification in terms of memory and execution time by leveraging time-triggered techniques for monitoring system events. Traditionally, runtime verification frameworks employ event-triggered monitors, where the invocation of the monitor occurs after every system event. Because systems events can be sporadic or bursty in nature, event-triggered monitoring behaviour is difficult to predict. Time-triggered monitors, on the other hand, periodically preempt and process system events, making monitoring behaviour predictable. However, software system state reconstruction is not guaranteed (i.e., missed state changes/events between samples). The first part of this thesis analyzes three heuristics that efficiently solve the NP-complete problem of minimizing the amount of memory required to store system state changes to guarantee accurate state reconstruction. The experimental results demonstrate that adopting near-optimal algorithms do not greatly change the memory consumption and execution time of monitored programs; hence, NP-completeness is likely not an obstacle for time-triggered runtime verification. The second part of this thesis introduces a novel runtime verification technique called hybrid runtime verification. Hybrid runtime verification enables the monitor to toggle between event- and time-triggered modes of operation. The aim of this approach is to reduce the overall runtime monitoring overhead with respect to execution time. Minimizing the execution time overhead by employing hybrid runtime verification is not in NP. An integer linear programming heuristic is formulated to determine near-optimal hybrid monitoring schemes. Experimental results show that the heuristic typically selects monitoring schemes that are equal to or better than naively selecting exclusively one operation mode for monitoring.en
dc.language.isoenen
dc.publisherUniversity of Waterlooen
dc.subjectsoftwareen
dc.subjectruntime verificationen
dc.subjectruntime monitoringen
dc.titleMethods for Reducing Monitoring Overhead in Runtime Verificationen
dc.typeMaster Thesisen
dc.pendingfalseen
dc.subject.programElectrical and Computer Engineeringen
uws-etd.degree.departmentElectrical and Computer Engineeringen
uws-etd.degreeMaster of Applied Scienceen
uws.typeOfResourceTexten
uws.peerReviewStatusUnrevieweden
uws.scholarLevelGraduateen


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record


UWSpace

University of Waterloo Library
200 University Avenue West
Waterloo, Ontario, Canada N2L 3G1
519 888 4883

All items in UWSpace are protected by copyright, with all rights reserved.

DSpace software

Service outages