【摘要】 算力互联网本质是在互联网体系架构上构建统一算力标识符，以算网云调度操作系统和高性能传输协议为基础，形成算力标准化、服务化的大市场和算力相互连接、灵活调用的逻辑网。当前算力互联网面临着算力资源利用率低和成本高等挑战。为解决这一问题，柔性计算应运而生，作为一种数据驱动、AI使能的新一代算力互联网资源调度范式。柔性计算通过精确的负载画像和服务质量（QoS）驱动的动态资源调度，优化算力资源池的利用率，在保障用户业务QoS的前提下显著降低用户的算力使用成本。深入研究了柔性计算的关键技术以及在算力互联网中的应用潜力，重点展示了其在提升算力利用效率和优化资源配置方面的创新优势。更多还原

【Abstract】 The Internet of computing resource was architected to establish a unified computing power identifier within the Internet framework. A computing-network-cloud scheduling operating system, integrated with high-performance transmission protocols, was utilized. Consequently, a large-scale, standardized, and service-oriented computing power market was formed, resulting in a logical network that enables the interconnection and flexible invocation of computing power. However, Internet of computing resource faces some important challenges such as low resource utilization and high costs. To solve these challenges, flexible computing has emerged as a new-generation resource scheduling paradigm that is data-driven and AI-enabled for Internet of computing resource. Through precise user profiling and quality of service(QoS)-driven dynamic resource scheduling, flexible computing can improve the resource utilization, significantly reducing the costs while ensuring QoS. The key technologies and applications of flexible computing are explored in Internet of computing resource, with the advantages in improving resource utilization and optimizing resource allocation highlighted.更多还原