Efficient and environment-friendly scheduling and resource management algorithms

Efficient and environment-friendly scheduling and resource management algorithms

Cloud computing has been very successful in providing on-demand computing resources for various compute-intensive applications. However, cloud-data-centers have increased in size to become large facilities with a huge environmental overhead and high carbon emissions. This research program aims at optimizing the scheduling of large-scale distributed applications with the primary target of decreasing power usage and carbon emissions. The two main directions being considered to achieve these objectives are: (1) focusing on the locality of computations and (2) recognizing that with edge resources the carbon content of power is complicated. To drive intelligent edge-note selection to reduce carbon emissions, this project proposes to create a more refined carbon-emissions model that captures information for specific load-serving entity (utility) contracts for edge sites (static, dynamic), as well as local renewables (eg. onsite solar), and energy storage (eg. onsite batteries). The ambitious goal of the project is then to design clever scheduling algorithms that will dynamically account for all these environmentally-driven changes.

Supporting mechanism: UChicago-CNRS PhD Joint Programme

Active dates: July 1, 2024-June 30, 2026