A Unified Communication Substrate for Monitoring and ODA-Driven Feedback and Control
Aug 27, 2026·
·
0 min read
Jim Brandt
Abstract
The continued growth in size, complexity, and technological diversity of compute systems, along with increasingly complex hosted workflows, creates an environment where both energy efficiency and throughput efficiency can decline for various reasons. This talk explores some of those reasons and discusses our current approaches that combine the scalable LDMS monitoring tool’s telemetry collection and flexible communication substrate with AI-driven data analyses to optimize resource utilization and workload throughput.
Authors
Jim Brandt
Senior Scientist
Jim Brandt is an affiliate of Sandia National Laboratories whose research interests include feedback-driven autonomous and efficient computing. His work includes the open-source Lightweight Distributed Metric Service (LDMS), used for monitoring extreme-scale platforms including El Capitan, and the AppSysFusion project, which integrates application and system performance and resource utilization data. Jim brings extensive experience in HPC monitoring, distributed services, platform-independent system design, data integration, APIs, and statistical and ML-based analysis of HPC monitoring data.