Massive energy consumption is an escalating threat to the environment. Large-scale computational grids consume a substantial amount of energy, and their energy requirements for powering and cooling are becoming comparable to the costs of acquisition. There is a lack of generally applicable methods for reducing energy consumption while ensuring good quality of service. This project will develop GridPac (Grid with Power-Aware Computing), a middleware environment that will allow grid managers and service providers to schedule multiple workflows across a distributed grid for system-wide optimization. GridPac will be based on a novel framework to support a variety of task-level workflows. The main features of this work are to develop:
(a) Novel static and dynamic algorithms for scheduling single and multiple workflows, which can be flexibly utilized by service providers in scalable grid environments.
(b) Control algorithms to account for dynamic adjustment of schedules using energy monitoring of the grid resources.
(c) Extensive benchmarking using a suite of commonly used grid workflows.
(d) A prototype middleware to assist IT organizations to better support their users while reducing energy costs.
The proposed work will lead to original scholarly contributions while harnessing the usage of computational grids. The project carries tremendous potential for economic, environmental, and societal impact. We will initiate new graduate and undergraduate level courses on related topics, and develop relevant tutorials, which will help to create awareness and educate a large audience on a critically important research topic.
Massive energy consumption is an escalating threat to the environment. Large-scale computational grids consume a substantial amount of energy, and their energy requirements for powering and cooling are becoming comparable to the costs of acquisition. There is a lack of generally applicable methods for reducing energy consumption while ensuring good quality of service. This project developed GridPac (Grid with Power-Aware Computing), a middleware environment that will allow grid managers and service providers to schedule multiple workflows across a distributed grid for system-wide optimization. This project involved two primary thrusts: 1) the creation of an infrastructure for use in evaluating the efficiency of computational grids; and 2) use of the resulting infrastructure to create new task-driven approaches to efficient computation in grids. The primary outcome of this project from Virginia Tech was integration of the PowerPack toolkit with grid environments; specifically, the UF grid toolkit. Our primary role in the project was to provide support for power measurement througout the project and ultimately create an interface for use with the UF grid toolkit. This interface was created and released on the http://scape.cs.vt.edu/gidpac website. Several graduate and undergraduate students were trained on the use and extension of the PowerPack toolkit as part of this project. The resulting tools and techniques can be used by grid researchers worldwide to improve the effiency of computational grids.
