Decentralized Caching in Two-layer Networks: Algorithms and Limits
Lin Zhang, Zhao Wang, Ming Bo Xiao, Gang Wu, Shaoqian Li · arXiv (Cornell University) · 2016
The decentralized caching is studied in two-layer networks, where users request contents through intermediate nodes (helpers) from a file server. By placing contents randomly and independently in each node and carefully designing the data delivery, the correlations of the pre-stored contents across layers can be utilized to reduce the transmission rates in the network. A hybrid caching scheme is developed by exploiting the cross-layer storage correlations, the single-layer multicast opportunities from the server (each helper) to helpers (the attached users), and the cross-layer multicast opportunities from the server to users. It is observed that, by the hybrid caching scheme, the achievable rate in the first layer is reduced without compromising the achievable rate in the second layer compared with the state of art. Furthermore, the achievable rate region is shown to be order-optimal and lies within constant margins to the information theoretic optimum. In particular, the multiplicative and additive factors are carefully sharpened to be $\frac{1}{48}$ and $4$, respectively.