High Performance and Scalable MPI Library for HPC and Deep Learning on Oracle HPC Cloud

Project

High Performance and Scalable MPI Library for HPC and Deep Learning on Oracle HPC Cloud

Principal Investigator

The Ohio State University

Oracle Principal Investigator

Sanjay Basu
Taylor Newill

Summary

The proposed research and development effort focus on two major directions to harness maximum performance and scalability for HPC and DL applications on Oracle HPC Cloud using the MVAPICH2 MPI libraries. These goals are achieved by redesigning and optimizing of MVAPICH2-X MPI library on Oracle HPC Cloud Instances. If successful, follow on research will help enabling deep learning environments and applications on upcoming Oracle GPU platforms with MVAPICH2-GDR, Horovod and DL frameworks like TensorFlow, PyTorch and MXNet.