Walker: DevOps Inspired Workflow for Experimentation

Abstract

Experimentation with computer networks under realistic conditions is a necessary step in debugging, profiling and validation towards real deployments and applications. Although the definition of relevant experimentation scenarios is usually relatively straightforward, their implementation and execution are unfortunately difficult and tedious. Generation of extensive experiment documentation assuring replicability is increasingly challenging even for experienced researchers. In this paper, we explain how a typical experimentation workflow can be supported using properly selected tools and components from the DevOps ecosystem, leading to repeatable, well-defined measurements. We start with a general approach using ad-hoc setups. Next, we show how the featured set of tools can be used with, and benefit from, existing testbeds.