Loading…
In-person + Virtual
November 6-9
Learn More and Register to Attend

The Sched app allows you to build your schedule but is not a substitute for your event registration. You must be registered for KubeCon + CloudNativeCon North America 2023 to participate in the sessions. If you have not registered but would like to join us, please go to the event registration page to purchase a registration.

Please note: This schedule is automatically displayed in Central Standard Time (UTC -6). To see the schedule in your preferred timezone, please select from the drop-down menu to the right, above "Filter by Date." The schedule is subject to change.
Back To Schedule
Thursday, November 9 • 11:00am - 11:35am
Environmentally Sustainable AI via Power-Aware Batch Scheduling - Atanas Atanasov, Intel & Daniel Wilson, Boston University

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

Tweet Share
Feedback form is now closed.
AI-training workloads running in batch environments like Kubeflow or Kueue define a growing percentage of datacenter power use. We demonstrate a solution that reduces these demands by implementing a batch extension and a pod-scheduling algorithm to define and minimize the power resources required for a workload. We achieve this by using modern k8s extension features to guide the scheduler using a non-linear hardware model of power as a function of utilization and a measurement of the current utilization of node components. In addition, we apply power limits to components on the system which are underutilized based on a model of the power requirements for the running jobs. We integrate the underlying components in a cloud-native batch scheduling framework and extend it with additional power-awareness capabilities and batch-job power control knobs for the user.
Speakers
avatar for Atanas Atanasov

Atanas Atanasov

Dr, Intel
Atanas has previous experience in the HPC Field as scientist and solution architect. Before Intel he was senior software engineer responsible for the implementation of a distributed rendering engine at Dassault-Systems 3DExcite and a Research Scientist at the Technical University... Read More →
avatar for Daniel Wilson

Daniel Wilson

Recent PhD Graduate, Boston University
Daniel Curtis Wilson received BS degrees in Computer Science and Computer Engineering from NC State University, Raleigh, North Carolina. He graduated with PhD degree in Computer Engineering at Boston University in August 2023. Prior to his current studies, Daniel worked at NetApp... Read More →

KubeCon23 Power Capping pdf
Thursday November 9, 2023 11:00am - 11:35am CST
W179 (Ground Level)
  Operations + Performance