The performance of MPI collectives communications is critical for scalable, high-performance computing. However, they are also a key limiting factor to parallel application efficiency and scalability. Offloading these communications to the network provides powerful capabilities for extending application scalability.
Join experts from Mellanox Technologies and Oak Ridge National Laboratory (ORNL) for a webinar discussing the results of a collaboration to find the solution that will address the MPI collectives communication scalability problem and enable the next generation of supercomputers to scale more efficiently.
The session will include a solution overview – featuring the latest technology developments - as well as a presentation of new benchmark results that show 20-35% reduction in job run-time for leading commercial HPC applications.