Optimizing monitoring of finite state properties through monitor compaction

Abstract

Runtime monitoring has proven effective in detecting property violations, but it can incur high overhead when monitoring just a single property -particularly when the property relates multiple objects. In practice developers will likely monitor multiple properties in the same execution which will lead to even higher overhead.This paper presents the first study of overhead arising during the simultaneous monitoring of multiple properties. We present two techniques for mitigating overhead in such cases that exploit reductions in cost that arise from sharing of information between property monitors. This sharing permits a single monitor to function in place of many monitors. We evaluate these techniques on the DaCapo benchmarks and 8 temporal correctness properties and find that they offer significant overhead reductions, as high as 57%, as the number of monitored properties increases.

RVprio: A tool for prioritizing runtime verification violations

“…All violations used in this paper came from an RV tool called JavaMOP [14]. Our choice is pragmatic-the violations that we use for training classifiers were generated from Java programs, JavaMOP is quite robust, easy to integrate with testing frameworks, publicly available, has been used in recent papers on performing RV during testing of open-source code [9][10][11], and is widely used in RV research [4,[19][20][21][22][23]. So, to evaluate RVPRIO on previously unseen violations, we also used JavaMOP to monitor the same properties from previous work on a new set of Java projects (Section 5).…”

BraceAssertion: Runtime Verification of Cyber-Physical Systems

“…This is similar to our approach; we aggregate repetitive atomic events to significantly reduce the number of events to be processed, and we use a global transition table backed by an efficient data structure. Our approach is also similar to [27], where inter-property and intra-property monitor compaction deal with large numbers of monitors and highfrequency events. Another way to improve the efficiency of runtime monitoring is to apply static analyses to eliminate unnecessary instrumentation [9].…”

Optimization Methods for Nop-Shadows Typestate Analysis

“…Purandare et al propose two novel approaches to mitigating monitoring overheads by monitor compaction [34]. The main idea of their techniques is to exploit over-lap, i.e., symbols shared among checked properties or common objects referenced by several property monitors, to synthesize supermonitors.…”