A New Approach and a Related Tool for Dependability Measurements on Distributed Systems

In recent years, experts in the field of dependability are recognizing experimental measurements as an attractive option for assessing distributed systems; contrary to simulation, measurement allows monitoring the real execution of a system in its real usage environment. However, the results of a recent survey have highlighted that the way measurements are carried out and measurement results are expressed is far from being in line with the approach commonly adopted by metrology. The scope of this paper is twofold. The first goal is to extend the discussion on the increasing role that measurements play in dependability and on the importance of cross-fertilization between the dependability and the instrumentation and measurement communities. The second objective is to present a different approach to dependability measurements, in line with the common practices in metrology. With regard to this, the paper presents a tool for dependability measurements in distributed systems that allows evaluating the uncertainty of measurement results. The tool is an enhancement of NekoStat, which is a powerful highly portable Java framework that allows analyzing distributed systems and algorithms. Together with the description of the tool and its innovative features, two experimental case studies are presented.