An instrumentation framework for the critical task of measurement collection in the future Internet

Abstract Experimental research on future Internet technologies involves observing multiple metrics at various distributed points of the networks under study. Collecting these measurements is often a tedious, repetitive and error prone task, be it in a testbed or in an uncontrolled field experiment. The relevant experimental data is usually scattered across multiple hosts in potentially different formats, and sometimes buried amongst a trove of other measurements, irrelevant to the current study. Collecting, selecting and formatting the useful measurements is a time-consuming and error-prone manual operation. In this paper, we present a conceptual Software-Defined Measurement (SDM) framework to facilitate this task. It includes a common representation for any type of experimental data, as well as the elements to process and collect the measurement samples and their associated metadata. We then present an implementation of this concept, which we built as a major extension and refactoring of the existing Orbit Measurement Library (OML). We outline its API, and how it can be used to instrument an experiment in only a few lines of code. We also evaluate the current implementation, and demonstrate that it efficiently allows measurement collection without interfering with the systems under observation.