Systems Lunch: Paul Townend, Autonomous Resource Management for Efficient and Sustainable Distributed Systems

Content

Image

Paul Townend

Speaker

Paul Townend (Umeå University)

Description

Modern distributed systems are composed of large-scale federations of interconnected clusters and data centers, forming the backbone for nearly all digital and AI-driven services. These systems deliver high performance and reliability, but at the cost of substantial deployment and management complexity, and significant energy usage, water consumption, and carbon footprint.

This talk presents techniques to mitigate and balance these factors through intelligent and autonomous resource management mechanisms. This is achieved by the temporal and spatial shifting of software workloads to improve application performance and system efficiency while reducing environmental impact, and is examined through the lens of Cloud and AI data centers. Real examples drawn from both industry and academia are used throughout.

I conclude by outlining a research vision for future green autonomous resource management approaches that can scale across massive geo-distributed Edge-Cloud systems and consider not only performance and sustainability, but also the broader impact of such systems on the communities in which they operate.