[简介] 智算业务的快速发展，计算资源需求急剧增长，对网络能力提出了更高标准。聚焦算力智联网，深入剖析数据入算、模型训练、模型推理三大典型智算业务场景需求，提出算力智联网关键技术要求。通过构建高效、灵活且可靠的智算基础设施，为智算业务提供差异化承载、灵活算网调度和无损传输等核心能力。此外，针对三大典型场景开展网络创新能力试点验证，显著提升了人工智能模型训练效率。

[简介]智算中心网络作为智算中心的连接底座，需要具备高性能、低时延的通信能力。智算中心网络体系是一个多要素融合的复杂系统，依赖于智算业务、机内/外交换芯片、网卡、网络设备等上下游产业协同创新。本文系统剖析了服务器/超节点内GPU互联网络、同园区机间互联网络、跨园区智算中心互联网络三大核心领域，探讨了智算网络的需求、挑战以及业界发展态势。中国移动创新提出OISA、GSE技术体系，以及弹性以太网聚合、精细化拥塞控制、物理层安全等多项创新技术，旨在构建超大规模、超高带宽、超低时延、超高可靠的智算中心网络，助力AI产业发展。

[Introduction] Due to the broadcast nature of wireless channels and the development of quantum computers, the confidentiality of wireless communication is seriously threatened. In this paper, we propose an integrated communications and security (ICAS) design to enhance communication security by using reconfigurable intelligent surfaces (RIS), in which the physical layer key generation (PLKG) rate and the data transmission rate are jointly considered. Specifically, to deal with the threat of eavesdropping attackers, we focus on studying the simultaneous transmission and key generation (STAG) by configuring the RIS phase shift. Firstly, we derive the key generation rate of the RIS-assisted PLKG and formulate the optimization problem. Then, considering the dynamic wireless environments, the optimization problem is modeled as a finite Markov decision process. We put forward a policy gradient-based proximal policy optimization (PPO) algorithm to optimize the continuous phase shift of the RIS, which improves the convergence stability and explores the security boundary of the RIS phase shift for STAG. The simulation results demonstrate that the proposed algorithm outperforms the benchmark method in convergence stability and system performance. By reasonably allocating the weight factors for the data transmission rate and the key generation rate, “one-time pad” communication can be achieved. The proposed method has about 90% performance improvement for “one-time pad” communication compared with the benchmark methods.

[Introduction] Along with the proliferating research interest in semantic communication (SemCom), joint source channel coding (JSCC) has dominated the attention due to the widely assumed existence in efficiently delivering information semantics. Nevertheless, this paper challenges the conventional JSCC paradigm and advocates for adopting separate source channel coding (SSCC) to enjoy a more underlying degree of freedom for optimization. We demonstrate that SSCC, after leveraging the strengths of the Large Language Model (LLM) for source coding and Error Correction Code Transformer (ECCT) complemented for channel coding, offers superior performance over JSCC. Our proposed framework also effectively highlights the compatibility challenges between SemCom approaches and digital communication systems, particularly concerning the resource costs associated with the transmission of high-precision floating point numbers. Through comprehensive evaluations, we establish that assisted by LLM-based compression and ECCT-enhanced error correction, SSCC remains a viable and effective solution for modern communication systems. In other words, separate source channel coding is still what we need.