Parsl & funcX Fest 2021 - The Parsl/funcX Community Meeting

Join us for the first Parsl & funcXFest Community Meeting. The meeting will be held as a virtual meeting in two half-day sessions on October 27 (9am-1pm CDT) and October 28 (1pm-5pm CDT), 2021.

The meeting will bring together researchers, developers, and cyberinfrastructure experts from around the country to discuss experiences using Parsl and funcX.

We are combining with the Parsl community to launch a joint community meeting. Parsl is a parallel programming library for Python.

Registration (free): https://forms.gle/yYW3AAk87uF7BTRS6. Please indicate if you are interested in presenting a lightning talk.

Agenda

Wednesday, October 27, 9 am - 1 pm CDT (14:00 - 18:00 UTC)

9:00 am - Welcome! - Ian Foster, University of Chicago/Argonne National Laboratory & Daniel Katz, University of Illinois at Urbana-Champaign. slides
9:10 am - Intro to Parsl and funcX - Kyle Chard, University of Chicago/Argonne National Laboratory. slides. video.
9:30 am - Session 1 (Chair: Ben Clifford)
  • Glen Hocky, Distributed prediction of protein structures with alphafold slides. video.
  • Matthew Feickert, Distributed statistical inference with pyhf powered by funcX slides. video.
  • Dante Domizzi Sánchez Gallegos, Sandbox apps and workflow:// schema in Parsl: design, implementation and application lifecycle slides. video.
  • Devesh Tiwari, Improving scalability of serverless functions on HPC clusters via enhancing funcX.
  • Eric Jonas, Scaling Shifts Sometimes Smashes Serverless Systems slides. video.
10:30 am - Tech talk 10:45 am - Break

11:15 am - Session 2 (Chair: Dan Katz)
  • Matthew Madany, A Multi-Datacenter Dataflow to Create the Most Detailed Image of the Brain
  • Zhengchun Liu, Rapid Deep Neural Network Training as a service. video.
  • Michael J Anderson, Parsl and funcX driven scientific workflows in HPC slides. video.
  • Joseph Moon, Memory-Unsafe Processes with Parsl slides. video.
  • Osama Almurshed, Neural layer as a Service Function on cloud–edge environment slides. video.
12:15 - Parallel Works Tech Talks
  • Matthew Shaxted, Parsl on Parallel Works for Science and Industry. slides. video.
  • Alvaro Vidal Torreira, Parsl on Elastic Cloud Clusters with JupyterLab. video.
12:45 - Tech talk
  • Kir Nagaitsev, Asynchronous APIs in funcX. slides. video.
13:00 - Day 1 Closing

Thursday, October 28, 1 pm - 5 pm CDT (18:00 - 22:00 UTC)

1:00 pm - Session 3 (Chair: Yadu Babuji)
  • Eliu Huerta, Accelerated, reproducible and scalable AI-driven gravitational wave discovery. slides video.
  • James Corbett, Parsl/Flux Integration slides. video.
  • Aristana Scourtas, A FAIR Approach to Data and Machine Learning Using FuncX. slides. video.
  • Luigi Marini, Creating Remote Imaging Pyramids for the Permafrost Discovery Gateway. slides. video.
  • Logan Ward, Colmena: Steering Ensemble Simulations on HPC. slides. video.
  • Tyler Skluzacek, A Serverless Framework for Distributed Bulk Metadata Extraction. slides. video.
2:10 pm - Tech Talk
  • Ben Clifford, Understanding workflow progress with Parsl monitoring. slides. video.
2:30 pm - Break

3:00 pm - Tech Talks
  • Douglas Thain, Resource Management for Dynamic Function Distribution. slides. video.
  • Ben Galewsky, Setting up kubernetes endpoints for FuncX. notebook. video.
  • Raf Vescovi, Globus Architecture for Data-Intensive Experimental Research. video.
4:00 - Session 4 (Chair: Ryan Chard)
  • Hemant Sharma, HEDM analysis using FuncX. video.
  • Matt Baughman, Automated, Cost-Aware FaaS in Heterogeneous Environments. slides. video.
  • Panos Patros, Rural AI: Using FuncX for Federated Learning of Hyperspectral Weed Detection Models. video.
  • Sierra Villarreal, Extreme Scale Survey Simulation with Python Workflows. slides. video.
  • Ved Kommalapati/Ananth Hariharan, EEPS (Efficiency Evaluator for Parallel Scripting) slides. video.
5:00 pm - Closing