Performance Modeling of Big Data-Oriented Architectures

Big Data applications provide new, disruptive tools to advance our knowledge about the mechanisms that characterize complex aspects of reality. Be it a high energy physics experiment or an analysis of social networks data, the strength of the approach is the availability of a huge richness of data; but, at the same time, it is also the main challenge, as this abundance of information must be processed at a bearable cost per information unit and requires higher scale systems to provide enough computing power. This is only possible if the Big Data platform is properly managed and exploited according to the needs of the applications, and a fundamental premise is the capability for a proper performance evaluation of the platform. In this chapter, we provide a glance over the main aspects of performance evaluation for Big Data architectures, together with some examples of model-based evaluation, in order to show how it is possible to characterize big scale architectures to support their correct management, and suggest a methodological coarse grain solution to exploit different conceptual and technical tools to integrate a flexible, model-based, performance analysis supported approach to Big Data systems design, capable of scaling up easily in the core evaluation stage means of Markovian agents.