X-ScaleSolutions will present a tutorial and live demo of their newest software offering, MVAPICH2-DPU MPI library. The tutorial, Accelerating HPC and AI Applications with MVAPICH2-DPU, X-ScaleHPL-DPU and X-ScaleAI-DPU Packages Live Demos, will give an overview of the MVAPICH2-DPU library. The MVAPICH2-DPU library takes advantage of the DPU features to offload communication components in the MPI library and accelerates HPC applications. It integrates key components enabling full computation and communication overlap, especially with non-blocking collectives. This tutorial will provide an overview of the MVAPICH2-DPU product, main features, and acceleration capabilities for a set of representative HPC and AI applications and benchmarks. Live demos of these applications will be shown to demonstrate the capabilities of the MVAPICH2-DPU MPI library. The tutorial will be led by Dr. Donglai Dai, Chief Engineer, Kyle Schaefer, Software Engineer.
In addition, Dr. Donglai Dai will give a talk titled Accelerating HPC and DL Applications using the MVAPICH2-DPU, X-ScaleHPL-DPU, and X-ScaleAI-DPU Packages. This talk will present an overview of the software architectures of MVAPICH2-DPU, X-ScaleAI and X-ScaleHPL-DPU products, discuss the underlying designs and benefits.
The MVAPICH2-DPU library takes advantage of the features to offload communication components in the MPI library and deliver best-in-class scale-up and scale-out performance for HPC and DL applications. It integrates key components enabling full computation and communication overlap, especially with non-blocking collectives.
The X-ScaleAI package provides a high-performance solution for distributed training for complex AI problems on modern HPC platforms including x86-64 and OpenPOWER with networks including InfiniBand, RoCE and NVLink. X-ScaleAI incorporates and optimizes components to support various Deep Learning Frameworks including TensorFlow, PyTorch, MXNet, and others with outstanding performance and scalability. DL applications, either CPU-based or GPU-based, may use X-ScaleAI to improve performance significantly. X-ScaleAI-DPU is a high performance solution to accelerate CPU-based distributed DNN training by utilizing the capabilities of data processing units (DPUs). X-ScaleAI and X-ScaleAI-DPU packages strive to achieve out-of-the-box optimal performance, one-click deployment and execution to avoid the struggle of DL stack setup for days or weeks.
X-ScaleHPL-DPU is enhanced to offload the communication to the NVIDIA BlueField DPU. The HPL-DPU can be run in two modes: Host mode and DPU mode with the MVAPICH2-DPU MPI library. In the DPU mode, communication offloading to DPU is enabled. In the Host mode, no such offloading occurs. This version of HPL is optimized in both Host mode and DPU mode.
This talk will present an overview of the software architectures of MVAPICH2-DPU, X-ScaleHPL-DPU, and X-ScaleAI-DPU products, discuss their underlying designs and benefits.
X-ScaleSolutions is also a sponsor of MUG 2023. Online attendance for the MUG 2023 conference is free if you would like to attend these talks.
For a trial license or to discuss our HPC and AI solutions, please contact X-ScaleSolutions. We are looking forward to sharing with the MUG community about X-ScaleSolutions and our ongoing enterprise solutions and research developments in high-performance computing, big data, deep learning and cloud computing.
About X-ScaleSolutions
X-ScaleSolutions specializes in a range of high-performance and scalable solutions for current generation systems ranging from small deployments of a couple systems to multi-petaflop systems and the emerging Exascale systems. The mission of the company is to develop innovative and leading edge software products, with a focus on four areas: 1) High-Performance Computing, 2) Deep Learning, 3) Big Data, and 4) Cloud Computing. To learn more about X-ScaleSolutions’ products and services, please visit www.x-scalesolutions.com.