Trace Checking of Metric Temporal Logic with Aggregating Modalities Using MapReduce

Abstract

Modern complex software systems produce a large amount of execution data, often stored in logs. These logs can be analyzed using trace checking techniques to check whether the system complies with its requirements specifications. Often these specifications express quantitative properties of the system, which include timing constraints as well as higher-level constraints on the occurrences of significant events, expressed using aggregate operators.In this paper we present an algorithm that exploits the MapReduce programming model to check specifications expressed in a metric temporal logic with aggregating modalities, over large execution traces. The algorithm exploits the structure of the formula to parallelize the evaluation, with a significant gain in time. We report on the assessment of the implementation-based on the Hadoop framework-of the proposed algorithm and comment on its scalability.

A Model-Driven Approach to Trace Checking of Pattern-Based Temporal Properties

“…Indeed, traces may contain a huge number of events, depending on the time span captured by the log, the nature of the system to which the log refers to (e.g., several virtual machines), the types of events monitored (e.g., high-level message passing events or low-level method calls) [10]. In the evaluation, we assess the relationship between the time taken to check a property on a trace and the structural properties of the trace (e.g., length, distribution of events) and the type of property to check; we also compare the performance of TEMPSY-CHECK with a state-of-the-art alternative technology.…”

“…The majority of the approaches proposed in this area -for example, [4], [34]- [36], including previous work of some of the authors [9], [10], [37] -focuses on the verification of temporal properties expressed using some temporal logic. These approaches define the trace checking/run-time verification problem in terms of a word problem, i.e., the problem of whether a given word is included in some languages, and rely on formal verification tools like model checkers or SAT/SMT solvers.…”

“…As for requirement R2, the trace checking procedure has to: 1) scale with respect to the length of the trace, because traces may contain a huge number of events, depending on the time span captured by the log, the nature of the system to which the log refers to (e.g., several virtual machines), and the types of events monitored (e.g., high-level message passing events or low-level method calls) [10]; 2) complete within practical time limits, because trace checking can be used not only for postmortem analysis, but also to enable real-time log analysis (as a complementary strategy to run-time monitoring), to promptly detect critical requirements violations.…”

CTL Model Checking in the Cloud Using MapReduce

“…To our knowledge, a few other techniques and tools have been introduced, with similar aims. E.g, [32] presents a MapReduce approach to check specifications expressed in a metric temporal logic over large execution traces, with aggregation modalities; [33] attempts (in a quite different context, i.e., Swarm Verification) to exploit massively parallel jobs running test randomization techniques to verify the correctness of mission critical software.…”

Efficient and Scalable Collection of Dynamic Metrics Using MapReduce

“…Geronimo et al [22] proposed a parallel genetic algorithm that uses MapReduce to automatically generate JUnit test suites. Bianculli et al [23] exploited the MapReduce framework to check specifications expressed in a metric temporal logic with aggregating modalities (over large execution traces).…”