Optimizing I/O in serial and parallel applications

A three-day tutorial on "Optimizing I/O" in serial and parallel applications will be hosted online from 12:30 PM CST - 2:30 PM CST on June 12 - June 14, 2023. The tutorial will include a high-level overview of the general strategies for optimizing I/O for improving code performance and will cover topics and examples related to parallel filesystems (Lustre), MPI I/O, and I/O in the context of checkpointing and AI. Hands-on exercises and demos will also be included in this tutorial. Below is the agenda for each day of the tutorial:

June 12, 2023, 12:30 PM CST - 2:30 PM CST: (1) Strategies and practices for reducing inefficient I/O in serial and parallel applications, including AI frameworks, and (2) Introduction to Parallel Filesystems (Lustre)

June 13, 2023, 12:30 PM CST - 2:30 PM CST: Reading and writing files in parallel using MPI

June 14, 2023, 12:30 PM CST - 2:30 PM CST: Checkpointing serial and parallel applications

The registration form for this tutorial is available at the following link: https://lnkd.in/g75Y_jdm . The registrations will close on June 9, 2023, at 5:00 PM CST. Meeting links will be shared with the registered participants.

The target audience of this tutorial includes those who are engaged in scientific software development but do not have prior experience with parallel filesystems, MPI I/O, and application-level checkpointing.

Prerequisites for this tutorial: Having familiarity with Linux, C/C++ programming and basic MPI will be helpful for the participants attending the first two days of the tutorial. Familiarity with Tensorflow and PyTorch will be useful for those attending the third day of the tutorial. Additionally, the participants will need to have active accounts on the HPC platforms of their interest if they need to run the examples or hands-on exercises covered in the tutorial.