Artificial Intelligence (AI) demands large data flows within datacenters, heavily relying on multicasting data transfers. As AI models scale, the requirement for high-bandwidth and low-latency networking compounds. The common use of electrical packet switching faces limitations due to its optical-electrical-optical conversion bottleneck. Optical switches, while bandwidth-agnostic and low-latency, suffer from having only unicast or non-scalable multicasting capability. This paper introduces an optical switching technique addressing the scalable multicasting challenge. Our approach enables arbitrarily programmable simultaneous unicast and multicast connectivity, eliminating the need for optical splitters that hinder scalability due to optical power loss. We use phase modulation in multiple planes, tailored to implement any multicast connectivity map. Using phase modulation enables wavelength selectivity on top of spatial selectivity, resulting in an optical switch that implements space-wavelength routing. We conducted simulations and experiments to validate our approach. Our results affirm the concept's feasibility and effectiveness, as a multicasting switch.

翻译：人工智能（AI）对数据中心内的大规模数据流提出了需求，尤其依赖多播数据传输。随着AI模型规模的扩大，对高带宽和低延迟网络的要求日益复杂。传统的电分组交换因其光电-电光转换瓶颈而面临局限性。光交换机虽具备带宽无关性和低延迟特性，却仅支持单播或不可扩展的多播功能。本文提出一种解决可扩展多播挑战的光交换技术。该方法能够实现任意可编程的同步单播与多播连接，无需使用因光功率损耗而阻碍可扩展性的光分路器。我们通过多平面相位调制技术，针对性地实现任意多播连接拓扑。利用相位调制在空间选择性基础上引入波长选择性，从而构建实现空间-波长路由的光交换机。通过仿真与实验验证，结果证实该概念作为多播交换机的可行性与有效性。