Wednesday, April 27th
|A-102: Standards (Architectures/Software/Tools Track)
Paper Title: A Standard API for Accessing Compute Engines on a Network
Paper Abstract: OpenSNAPI, a project of the Unified Communication Consortium (UCF, see ucfconsortium.org), is a proposed standard application programming interface (API) for accessing compute engines on a network. The OpenSNAPI unified API is intended to expand the applicability of emerging use cases, such as in-network computing, and ultimately enabling more efficient data processing through broader deployments of SmartNICs. A proof-of-concept implementation of the API based on the OpenSNAPI software stack├é┬ is now available for fast compute migration. This API could be used to dispatch user functions from a host CPU to a SmartNIC (DPU) or remote servers. It can also meet the needs of large-scale irregular applications (such as semantic graph analysis) composed of many coordinating tasks operating on a data set so big that it must be stored on many physical devices.
Paper Author: Pavel Shamis, Sr Principal Research Engineer, Arm
Author Bio: Pavel Shamis (Pasha) is a Senior Principal Research Engineer at Arm. His work focuses on co-design software, hardware building blocks for high-performance interconnect technologies, development of communications middleware, and novel programming models. Before joining Arm, he was a Research Scientist at Oak Ridge National Laboratory (ORNL) where he worked on high-performance communications domains including Collective Communication Offload (CORE-Direct & Cheetah), OpenSHMEM, and Open Unified Communication X (Open UCX). Before joining ORNL, Pavel worked at Mellanox Technologies, where he led the HPC team and was a key driver in enabling the Mellanox HPC software stack. He holds multiple patents dealing with in-network accelerators. Pavel received a 2015 R&D 100 award for his work on the CORE-Direct in-network computing technology and the 2019 R&D 100 award for developing the Open UCX software framework for HPC, storage, data analytics, and AI.