From Measures to Conclusions Using Analytic Hierarchy Process in Dependability Benchmarking

Dependability benchmarks are aimed at comparing and selecting alternatives in application domains where faulty conditions are present. However, despite its importance and intrinsic complexity, a rigorous decision process has not been defined yet. As a result, benchmark conclusions may vary from one evaluator to another, and often, that process is vague and hard to follow, or even nonexistent. This situation affects the repeatability and reproducibility of that analysis process, making difficult the cross-comparison of results between works. To mitigate these problems, this paper proposes the integration of the analytic hierarchy process (AHP), a widely used multicriteria decision-making technique, within dependability benchmarks. In addition, an assisted pairwise comparison approach is proposed to automate those aspects of AHP that rely on judgmental comparisons, thus granting consistent, repeatable, and reproducible conclusions. Results from a dependability benchmark for wireless sensor networks are used to illustrate and validate the proposed approach.