Evaluating the Performance and Energy Consumption of Distributed Data Management Systems

In the last couple of years it became more difficult for end users of data management system to optimize existing systems for performance and energy consumption. The reasons are the big number of data management systems to chose from, extensive use cases and a variety of hardware configurations. The number of factors that influence the performance and energy efficiency rises even more when it comes to distributed data management systems. For instance, in addition to vertical scale-out effects the horizontal scale-out ones have to be considered. This paper introduces an enhanced version of a Queued Petri Nets model whose simulation runs allow to predict and to study the performance and energy consumption of a distributed data management system. In contrast to traditional ways, the simulation runs can reduce both investments in time and hardware. The model's prediction in terms of performance and energy consumption were evaluated and compared with the actual experimental results. The predicted and experimental results for the response times differ by nearly 20 percent on average.