Moving contents proximity to the network edge and proactively caching popular contents at multiple infrastructures are promising directions for solving the backhaul congestion problem. This letter proposes and evaluates a multi-level cache-enabled network, where cache-hit users can fetch their data from the available cache at either small base-stations, unmanned aerial vehicles, or cache-enabled mobile device-to-device users. Cache-miss users, on the other hand, fetch their data from the central cloud via limited capacity backhaul links. This letter considers the problem of maximizing the network weighted-sum rate by jointly determining the users’ mode of operation and their transmit powers, subject to backhaul capacity and transmit power constraints. After showing how the association problem can be formulated as a generalized assignment problem, the letter solves the transmit power problem using an iterative function evaluation apprach. The resulting mode-selection and power-allocation (MSPA) iterative algorithm is then tested through numerical simulations, which suggest that, while being easily implementable, the proposed multi-level caching can substantially relieve the backhaul congestion, especially in dense-networks, and at low-backhaul capacity regimes.

中文翻译：

---

多级缓存网络中的模式选择和功率分配

将内容靠近网络边缘并在多个基础设施中主动缓存流行内容是解决回程拥塞问题的有希望的方向。这封信提议并评估了一个启用多级缓存的网络，其中缓存命中的用户可以从小型基站、无人机或启用缓存的移动设备到设备用户的可用缓存中获取他们的数据。另一方面，缓存未命中用户通过有限容量的回程链路从中央云获取数据。这封信考虑了通过联合确定用户的操作模式及其发射功率来最大化网络加权和速率的问题，受回程容量和发射功率约束。在展示了如何将关联问题表述为广义分配问题之后，这封信使用迭代函数评估方法解决了发射功率问题。然后通过数值模拟测试产生的模式选择和功率分配 (MSPA) 迭代算法，这表明，虽然易于实现，但所提出的多级缓存可以显着缓解回程拥塞，尤其是在密集网络中，并且在低回程容量的情况下。