摘要
AI智算基础设施是实现人工智能快速发展的关键要素,电信运营商在构建AI智算基础设施时需要部署合理的技术架构,为用户提供高性能的AI计算能力。文章介绍了AI智算基础设施中AI-IaaS层、AI-PaaS层以及AISaaS层主要包括的组件以及主要功能。通过对当前AI智算基础设施能力短板进行分析,得出了运营商在部署新一代AI计算基础设施架构时,应重点优化的五项关键能力:高性能并行计算能力、兼容异构的开放能力、可信计算能力、资源弹性调度能力以及低碳绿色计算能力。
AI computing infrastructure is the key element for the rapid development of AI.Telecom operators need to build advanced AI computing infrastructure to provide users with high-performance AI computing capabilities.This paper introduces the main components and architecture AI-IaaS layer,AI-PaaS layer and AI-SaaS layer.By analyzing the shortness of current AI intelligent computing infrastructure,five key capabilities are recommended when deploying the new generation of AI computing infrastructure:supporting high-performance parallel computing capabilities,compatible heterogeneous computing capabilities,trusted computing capabilities,virtual elastic scheduling capabilities,and low-carbon green computing capabilities.
作者
王鑫
赵慧英
汪悦
李贝
郭熹
Wang Xin;Zhao Huiying;Wang Yue;Li Bei;Guo Xi(China Unicom Research Institute,Beijing 100176,China)
出处
《信息通信技术》
2023年第1期56-63,共8页
Information and communications Technologies
关键词
人工智能
基础设施
AI芯片
智算
低碳
Artificial Intelligence
Infrastructure
AI Chipset
Computing
Low-carbon
