Affiliation:
1. Technion—Israel Institute of Technology
2. University of Texas at Austin
Abstract
Despite the popularity of GPUs in high-performance and scientific computing, and despite increasingly general-purpose hardware capabilities, the use of GPUs in network servers or distributed systems poses significant challenges.
GPUnet is a native GPU networking layer that provides a socket abstraction and high-level networking APIs for GPU programs. We use GPUnet to streamline the development of high-performance, distributed applications like in-GPU-memory MapReduce and a new class of low-latency, high-throughput GPU-native network services such as a face verification server.
Funder
National Science Foundation
Israel Science Foundation
Israeli Ministry of Science
Publisher
Association for Computing Machinery (ACM)
Cited by
52 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献
1. CARE: Cloudified Android With Optimized Rendering Platform;IEEE Transactions on Multimedia;2024
2. rShare: Alleviating long startup on the Cloud-rendering platform through de-systemization;Journal of Systems Architecture;2023-12
3. GPU Graph Processing on CXL-Based Microsecond-Latency External Memory;Proceedings of the SC '23 Workshops of The International Conference on High Performance Computing, Network, Storage, and Analysis;2023-11-12
4. CPU-free Computing: A Vision with a Blueprint;Proceedings of the 19th Workshop on Hot Topics in Operating Systems;2023-06-22
5. Multi-GPU Communication Schemes for Iterative Solvers: When CPUs are Not in Charge;Proceedings of the 37th International Conference on Supercomputing;2023-06-21