SubMIT Users Meeting - operated by Physics Basic Computing Services

Tuesday 22 Apr 2025, 10:00 → 11:00 America/New_York

Marlar Lounge (37-252) (MIT)

The SubMIT analysis facility is for anyone in the Physics Department and is a set of servers that provide interactive access to substantial storage at high speeds, enabling sophisticated data analyses with very fast turnaround times. Additionally, it seamlessly integrates massive processing resources for large-scale tasks by connecting to a set of powerful batch processing systems.

More information about SubMIT: https://submit.mit.edu/

Zoom: https://mit.zoom.us/j/96743699673?pwd=b3h2Q3c3cVQwYW12blhMUG5SWXZCZz09

10:00 → 10:10 SubMIT Overview & News

Speaker: Matthew Heine (Massachusetts Institute of Technology)

UserMeetingApril2025.pdf

10:10 → 10:25 Batch computing on SubMIT

Luca will discuss setting up and managing batch jobs using Slurm and HTCondor on the subMIT infrastructure. He includes basic job submission, monitoring, and troubleshooting, as well as more complex workflows involving Python scripts and Singularity containers. Participants will find an entry point to efficiently execute and manage batch jobs on SubMIT.

Speaker: Luca Lavezzo (MIt)

10:25 → 10:55 Modeling gravitational waves using highly distributed parallel computing

Professor Scott Hughes will discuss recent work modeling gravitational wave sources using large-scale parallel computing. His research focuses on solving complex differential equations that describe astrophysical phenomena responsible for generating gravitational waves, such as extreme mass ratio inspirals and other compact object systems. These calculations require an enormous number of repeated computations—on the order of hundreds of millions of equation solves—which are addressed by leveraging highly distributed, high-throughput computing resources. Professor Hughes will describe how the subMIT platform and the Open Science Grid (OSG) have enabled his team to dramatically accelerate these computations, reducing run times from months to days. The talk will cover the astrophysical motivation for this work, the computational challenges involved, and how scalable workflows and parallelization strategies are advancing this research. He will also outline current efforts and future plans to further optimize workflows for even more demanding simulations.

Speaker: Scott Hughes (Massachusetts Institute of Technology)

10:55 → 11:00 Roundtable & Discussion

Speakers: Amer Al-Hiyasat (MIT), Hans Moritz Guenther (Massachusetts Institute of Technology), Molly Park (Massachusetts Institute of Technology), Prajwal Mohan Murthy (MIT LNS), Yin Lin (Massachusetts Institute of Technology)