Coded caching (CC) exploits cumulative cache memory at user devices and coding to transform unicast traffic into multicast transmissions. While information theoretic results show significant gains over uncoded caching for various network topologies, its practical benefits remain unclear. In this work, we investigate CC for on-demand video streaming over large wireless local area networks, where multiple users are served simultaneously by spatially distributed access points. Users asynchronously request video chunks from a content library. We propose a decentralized, asynchronous, and location-independent cache placement scheme combined with an "over IP" delivery mechanism operating at higher network layers, leaving the physical and MAC layers unchanged. For this scheme, we characterize the achievable goodput region, where goodput is defined as the number of video chunks per unit time delivered to users' playback buffers, and formulate the corresponding fairness problem as a convex maximization. We develop a dynamic scheduling algorithm that provably achieves the optimal fairness point under stationary conditions with reduced complexity, and introduce a heuristic to further lower complexity. Numerical results demonstrate significant gains over baseline schemes, including conventional prefix caching, orthogonal sub-channel allocation with spatial reuse, and a CSMA-inspired distributed coordination approach, showing that CC can be implemented as a scalable and compatible over IP solution for existing WLANs.

翻译：缓存编码利用用户设备上的累积缓存内存和编码技术，将单播流量转化为组播传输。尽管信息论结果表明，对于各种网络拓扑，缓存编码相较于未编码缓存具有显著优势，但其实际效益仍不明确。本文研究用于大规模无线局域网上点播视频流的缓存编码，其中多个用户由空间分布的接入点同时提供服务。用户异步从内容库请求视频块。我们提出一种去中心化、异步且位置无关的缓存部署方案，结合一种运行在更高网络层、基于“over IP”的交付机制，保持物理层和MAC层不变。针对该方案，我们刻画了可实现的有效吞吐量区域（其中有效吞吐量定义为单位时间传递到用户播放缓冲区的视频块数量），并将相应的公平性问题表述为凸优化。我们开发了一种动态调度算法，在平稳条件下以降低的复杂度证明可实现最优公平点，并提出一种启发式方法以进一步降低复杂度。数值结果表明，与传统前缀缓存、具有空间复用的正交子信道分配以及受CSMA启发的分布式协调方法等基准方案相比，该方案具有显著优势，表明缓存编码可作为现有无线局域网中一种可扩展且兼容的over IP解决方案实施。