当前位置:首页 > 报告详情

以太网不断发展借助超级以太网联盟为人工智能的未来提供动力.pdf

上传人: a****d 编号:185000 2024-10-07 21页 3.91MB

1、1|2024 SNIA.All Rights Reserved.AMD Official Use Only-GeneralEthernet Evolved:AIs Future with the Ultra Ethernet Consortium(UEC)Presented by:J Metz,Ph.DChair,UECJ Metz,Ph.D,2|2024 SNIA.All Rights Reserved.AMD Official Use Only-General“Ultra Ethernet?”-Agenda Why“Ultra Ethernet?”Who is Ultra Ethernet

2、 Consortium(UEC)?Philosophy of UEC Overview of projects Where to get more information23|2024 SNIA.All Rights Reserved.AMD Official Use Only-GeneralAI for Networking,or Networking for AI?Many articles/blogs have talked about how AI can change the networking infrastructure but what network infrastruct

3、ure do you need to have enough AI to change the networking infrastructure?Is it more than just superfast speeds and feeds?Massive data sets,parallel processing requirements Where does the data need to be,and when?4|2024 SNIA.All Rights Reserved.AMD Official Use Only-GeneralThe AI Monster AI workload

4、s need Ever-increasing Memory Bandwidth Ever-increasing Memory Capacity(Near)Instantaneous Data Access(Exabytes)Intermittent data surges”Straggler”data(tail latency)significantly impacts completion time Extended operation duration(hours,days)5|2024 SNIA.All Rights Reserved.AMD Official Use Only-Gene

5、ralWhich Network?UEC=Scale-Out6|2024 SNIA.All Rights Reserved.AMD Official Use Only-GeneralAI Problems To Solve Memory Bandwidth vs.Capacity vs.Latency Computation-Bound Workloads E.g.,Summarization:processes large input context simultaneously(parallel)Higher weights reusability Memory-Bound Workloa

6、ds E.g.,Generation:produces single word at a time(iteration)Lower weight reusability Recommendation Workloads spend almost 60%of time in Network I/O*Meta,OCP 2022 Global Summit7|2024 SNIA.All Rights Reserved.AMD Official Use Only-GeneralThe Goodput Dilemma Logarithmic expansion of the relationships

word格式文档无特别注明外均可编辑修改,预览文件经过压缩,下载原文更清晰!
三个皮匠报告文库所有资源均是客户上传分享,仅供网友学习交流,未经上传用户书面授权,请勿作商用。
The data provided outlines a presentation by the Ultra Ethernet Consortium (UEC) on the evolution of Ethernet for AI applications. Key points include: - The need for "Ultra Ethernet" arises from the requirements of AI workloads, which demand high-speed data transfer, massive data processing, and low-latency performance. - The UEC aims to address these needs by developing a new network infrastructure that is scalable, cost-effective, and capable of supporting the demands of AI and high-performance computing (HPC). - The presentation highlights the challenges in networking for AI, such as data ordering, congestion control, and the impact of tail latency on computation time. - UEC's approach includes flexible packet ordering, advanced security features, and improved network telemetry to enhance network visibility and performance. - The consortium has grown rapidly, with over 70 companies and 800+ volunteers contributing to develop standards, protocols, and specifications for the new network technology. - The proposed Ultra Ethernet network is intended to be backward compatible, using libfabric as its north-bound API, and to integrate with existing frameworks. - It aims to achieve a balance between network utilization and tail latency, while ensuring security and supporting modern APIs. For more information, one can visit the UEC's website at [https://ultraethernet.org/](https://ultraethernet.org/).
"Ultra Ethernet如何改变网络基础设施?" "AI与Ethernet Evolved:AI的未来有何关联?" "Ultra Ethernet Consortium (UEC)的目标和项目有哪些?"
客服
商务合作
小程序
服务号
折叠