Join us at the CARLA Latin America High Performance Computing Conference for the keynote presentation by Kate Keahey, Scientist at Argonne National Laboratory and University of Chicago. Kate will deliver her keynote, “Infrastructure for New Ideas,” at 9:10 AM on Thursday in the Talk of the Town venue. Don’t miss this opportunity to learn about building scientific instruments for computer science innovation and hear insights from the leader of the Chameleon project.

https://carlaconference.org/program/

See below for the full abstract of Kate Keahey's talk:

We live in interesting times where new ideas in computing and technology emerge at ever increasing rates in AI, edge computing and IoT, and programmable networking to name just a few. These innovations open up unprecedented opportunities for all kinds of scientific progress from biotechnology to engineering. Given their critical role, the question arises who creates opportunities for computer science? How can we create a scientific instrument where computer science ideas can be tried, tested, and adopted or discarded? What would such a scientific instrument look like and would it engage with its community? How would such instrument evolve to follow the evolution of science? And lastly, how would it negotiate transition from innovation to mainstream adoption?

In this talk, I will talk about how to build this type of scientific instrument supporting the exploration of new ideas in the cyberinfrastructure space. I will share the insights, design strategy, and the lessons learned from building and operating the Chameleon computer science research platform through the last decade. We will take the journey from a base cloud testbed design and track its evolution through diversifying its hardware to support innovative architectures (e.g., Fugaku nodes), accelerators, disaggregated hardware (Liqid, GigaIO), and ultimately taking it form the datacenter and into the field by introducing support for edge hardware based on single board computers (Raspberry Pis and NVIDIA nanons). We will see how the platform evolved to support emergent ideas coming from its by now 13,000 strong user community at a reasonable cost, and allowing it to support a significant scientific output of 800+ publications. Lastly, I will share stories of research and education projects in the edge to cloud continuum and discuss their impact both on science and on scientific sharing through reproducible digital artifacts.