This document describes the performance methods and tools to predict the execution time of big data applications that have been developed during the first year of EUBra-BIGSEA. Moreover, this report introduces the optimization models which will be the core of horizontal elasticity policies. Optimization-based policies will trigger the configuration of the cloud infrastructure providing QoS guarantees for big data applications execution while minimising resource usage costs.
The ultimate goal of the optimization based policies that will be developed within EUBra-BIGSEA is to determine the system configuration leading to the minimum cost, embedding performance prediction methods in an optimization framework. Performance models will be kept alive at run-time, and will be integrated within an optimization algorithm, to provide the proactive-based approach with the insights to perform the dynamic adjustment of the system configuration.