The ACM International Symposium on High-Performance Parallel and Distributed Computing (HPDC) is an academic research conference, started in 1992, that focuses on topics around systems software and middleware layer of high-performance and distributed computing. It’s historically been attended by academic researchers (CS faculty, grad students) with some DOE research attendees and limited industry presence. The conference has historically leaned towards job scheduling, workflow management, and distributed runtimes.
I was invited to HPDC 2026 in Cleveland, OH for two reasons:
- Provide a keynote to the PERMAVOST workshop
- Contribute a short talk to the Moonshot Ideas session
PERMAVOST keynote
PERMAVOST is the Performance EngineeRing, Modeling, Analysis, and VisualizatiOn STrategy workshop. I am thinking of talking about…
Thought 0
What do the I/O and infrastructure experts of tomorrow need to understand about the world of AI? What wouldn’t they have learned about what’s happening in the state of the art? What will hit them in the face hard as they enter the workforce or are thrust into a fast-paced, high-visibility AI project?
Thought 1
Performance tools measure machine speed, not scientific progress. That’s been fine in the case where the model is fundamental and the role of the machine is to use a model to generate data. Observability was easy; you measure how fast a system is performing, and you have a good proxy for how productive that system is.
In the world of data-driven discovery though, the model doesn’t exist a priori, and the results may come from a model or system that may perform well but generate poor or unpredictable results. This makes good observability existential; getting observability wrong is no longer just about getting to the right answer faster, it’s about getting to the right answer period.
Moonshot Ideas session
Still working on this.