Software Metrics for Green Parallel Computing of Big Data Systems

Abstract

Big Data is typically organized around a distributed file system on top of which the parallel algorithms can be executed for realizing the Big Data analytics. In general, the parallel algorithms can be mapped in different alternative ways to the computing platform. Hereby each alternative will perform differently with respect to the environmentally relevant parameters such as energy and power consumption. Existing studies on deployment of parallel computing algorithms have mainly focused on addressing general computing metrics such as speedup with respect to serial computing and efficiency of the use of the computing nodes. In this paper, we report on the elicitation of green metrics for big data systems that are required when analyzing deployment alternatives. To this end we use the existing systematic literature reviews and identify, and discuss the important green computing metrics for big data systems.