CNS Core: Small: Language Runtime Support for Energy- Aware Applications

As energy consumption becomes a unifying concern for modern computing platforms, concerted efforts on energy optimization are required from all layers of the computing stack. Lower-layer solutions offer innovative designs on hardware and systems software to improve energy efficiency, whereas higher-layer solutions rely on innovations over programming models and application software design. Unfortunately, few solutions exist to bridge the gap in between, leading to missed opportunities for lower-layer solutions and limited deployment of higher-layer solutions. The goal of this project is to streamline the higher-layer and lower-layer solutions through innovations at a crossroads where the logical world of the application and the physical world of the systems meet, namely, language runtimes. The intellectual merit of the project lies in a number of runtime-centric cross-layer energy optimization solutions that judiciously draw from both higher-level application information and lower-level system information. Concretely, this project will address two fundamental problems in streamlining application-level energy management with underlying systems: (1) interference: how application-level energy management can remain effective in the presence of complex multi-threading and multi-processing, and (2) uncertainty: how solutions from different layers can compose in an environment where each layer may introduce its own share of uncertainty. The highlights of the solutions include a bi-directional energy mediator to gracefully mediate the interference from threads and processes, and a stochastic energy optimizer service to continuously optimize energy consumption in the presence of uncertainty. The broader impacts of the project constitute several dimensions. First, studying energy optimization at the level of managed language runtimes has the potential to impact a majority of computing platforms we use today, ranging from Android-enabled smartphones, to graph processing engines deployed on data centers, to Javascript-powered web applications. Second, the cross-layer nature of this project will bring researchers from different communities together, through workshops and seminars. Third, the project presents research opportunities for underrepresented students, and will lead to curriculum enrichment at the frontier of computing. The project will yield a number of algorithms, compiler implementations, managed runtime implementations, and tools that form the ecosystem of an energy-aware language runtime. All software artifacts produced from the project will be open-source, and their repositories - together with all experimental data and publications - will be made available at http://www.cs.binghamton.edu/~davidl/GreenRuntime throughout the duration of the project, and continuously maintained for an additional 3 years after the completion of the project. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.