This paper proposes a point-to-point passive optical network (P2P-PON) architecture as an energy-efficient and low-latency backhaul solution for visible light communication (VLC)-enabled indoor fog computing systems. The proposed architecture passively interconnects VLC access points and distributed in-building fog servers through dedicated optical links, enabling flexible peer-to-peer connectivity and efficient traffic aggregation. A mixed integer linear programming (MILP) framework is developed to jointly optimize processing resource allocation, traffic routing, power consumption, and end-to-end queuing delay across a multi-layer fog computing infrastructure. The model explicitly captures the power consumption of both networking and processing elements and incorporates a piecewise linear approximation of an M/M/1 queuing model to represent delay-sensitive applications. The performance of the proposed P2P-PON architecture is evaluated and compared against an arrayed waveguide grating router (AWGR)-based PON architecture under multiple indoor traffic scenarios. The results show that the proposed P2P-PON architecture reduces total power consumption by up to 64\% under power-aware optimization and by 15\% under delay-aware optimization, while reducing average end-to-end queuing delay by up to 76\% compared to the AWGR-PON architecture, due to the improved in-building connectivity and more effective utilization of distributed fog resources.

翻译：本文提出一种点对点无源光网络（P2P-PON）架构，作为可见光通信（VLC）使能的室内雾计算系统的高能效、低延迟回传解决方案。该架构通过专用光链路无源互连VLC接入点与分布式楼内雾服务器，实现灵活的对等连接与高效的流量汇聚。研究建立了混合整数线性规划（MILP）框架，用于联合优化多层雾计算基础设施中的处理资源分配、流量路由、功耗及端到端排队延迟。该模型精确刻画了网络设备与处理单元的功耗特性，并采用分段线性化的M/M/1排队模型来表征时延敏感型应用。通过多种室内流量场景的仿真实验，将所提P2P-PON架构与基于阵列波导光栅路由器（AWGR）的PON架构进行性能对比。结果表明：在功耗感知优化模式下，P2P-PON架构总功耗最高可降低64%；在时延感知优化模式下，总功耗可降低15%；得益于增强的楼内连接性与分布式雾资源的高效利用，其平均端到端排队延迟较AWGR-PON架构最高可减少76%。