Characterization of a Workload Generator for Content-based Publish-subscribe Research Evaluation

Achieving large-scale content-based publish-subscribe has been an ambitious research agenda that has received tremendous attention over the last two decades by researchers from distributed systems and networking communities.Simulations have been the most common approach for evaluating solutions.However, the research community on this topic neither has shared workload assumptions nor standard workload generation methodologies.As a result, each effort has introduced its own assumptions and ad-hoc workload generation methodologies.Also, comparison to related alternatives has often been neglected.This has made it difficult to understand the performance gains of one contribution over related alternatives.This paper reports an effort to enhance a workload generation tool for content-based publish-subscribe research using Google groups data.It is enhanced with a visual characterization of the generated workload, given a set of parameters.The workload generated can be characterized in terms of popularity and locality.The resulting software contributes to generating well-specified workloads, facilitates experiment reproducibility, and it will also be time-saving in evaluation processes.