Approximating quality contracts for energy auto-tuning software

Abstract

An emerging trend for future software systems is self-optimization, especially w.r.t. energy efficiency. Models of soft- and hardware components at runtime, expressing current and alternative system configurations, are exploited to improve service utility as well as to decrease energy consumption. In recent work we showed how quality contracts - expressing dependencies between software and hardware components - can be used for energy auto-tuning. Notably, the declared provisions and requirements of individual components depend on software containers (i.e., the servers, components are deployed on) and thus, cannot be declared completely at design time. In this paper we present a semi-automated contract creation process that combines manual created contract templates with benchmarking and mathematical approximations for nonfunctional properties depending on the components' runtime behavior as well as their service's input parameters. We identify individual process activities and show how the process can be applied to approximate the nonfunctional behavior of software components providing simple sorting functionality.