Communication in Distributed Architecture

Design and implementation of data transfer and network communication solutions targeted at the requirements of simulation tools. The course will be broken into three parts:

1. Review of current message passing protocols and implementations, such as MPI
2. An implementation appropriate for Stage2;
3. A demonstration of the solution's capabilities within Stage2, including performance characterization.

Course Weighting

The primary course outcome is to present a software design for a network-based communication system that includes good practice in documentation, testing, and presentation of results. Considering this and the quality of the work, the grades for the course will be weighted as follows:

• 40% Documenting the communication design within Stage 2 including within its repositories.
• 30% Preparation and documentation of a working example of the communication system.
• 20% Thirty-minute presentation of the system’s design and integration within Stage 2 .
• 10% Timely updates on progress via regular meetings.

Resources

Message-passing interfaces and protocols used in scientific software:

• open-mpi.org - Industry-standard implementation of the message-passing interface (MPI) protocol
• OpenMP-API-Specification-6-0 - Heterogeneous computing API

Tools:

• Paraview - open-source software framework for simulations with in situ visualization and analysis capabilities
• IOuring - Efficient IO with io_uring
• liburing - Library providing helpers for the Linux kernel io_uring support