Zoom: https://illinois.zoom.us/j/85794614413?pwd=QaoY4Q0XgzG4raCUOKJ1AzvrrndtMc.1
Abstract:
The computing landscape is undergoing a profound shift. Growing demand, climate-driven volatility, and tight resource constraints are putting pressure on long-held assumptions of reliable and abundant energy. From datacenters to edge devices, computer systems now operate in environments where energy supply, thermal conditions, and grid dynamics are unpredictable and deeply intertwined. In this talk, I will present my work on redesigning system abstractions to navigate this new reality. I will focus on how enabling dynamic and integrated resource management --- across energy, thermal, and compute resources -- can improve reliability and efficiency under such constraints.
First, I will introduce Ecovisor, a virtualization layer that makes energy signals such as price, power availability, carbon intensity, and grid reliability first-class citizens in the system stack, allowing applications and system software to respond to grid dynamics in real time. Next, I will describe how these principles extend to thermal management in ectothermic systems, where jointly managing compute and heat enables systems to balance reliability, availability, and efficiency amid harsh and variable environmental conditions. Together, these approaches demonstrate how rethinking traditional system abstractions can enable computer systems to maintain reliable and sustainable operation even when built on inherently unreliable and dynamic substrates.
Finally, I will outline my vision for future systems that move beyond managing individual resources to address the interdependencies and couplings across them -- from energy and thermal dynamics within computing environments to integration with societal-scale infrastructures such as grids, buildings, and transportation -- ensuring resilient, adaptive, and sustainable computing in an increasingly interconnected world.
Bio:
Noman Bashir is a Computing & Climate Impact Postdoctoral Fellow at the MIT Climate & Sustainability Consortium (MCSC) and Computer Science & Artificial Intelligence Laboratory (CSAIL). His research reimagines the design and operation of computer systems to address rising demand, climate-driven volatility, and growing interdependence with societal infrastructure. Noman's work spans virtualizing energy systems for grid-responsive computing, developing ectothermic systems that co-manage compute and thermal resources, and advancing uncertainty-aware modeling and control in dynamic environments. His research has been deployed in hyperscale datacenters, including at Google, and recognized with awards at ACM SIGMETRICS, e-Energy, Supercomputing, and other venues. Noman earned his Ph.D. from the University of Massachusetts Amherst, and his current work aims to enable resilient and sustainable computing systems that integrate seamlessly with critical societal infrastructures. To know more about his work, please visit https://noman-bashir.github.io/.
Faculty Host: Klara Nahrstedt
Meeting ID: 875 9461 4413; Password: csillinois