This paper considers a hierarchical caching system where a server connects with multiple mirror sites, each connecting with a distinct set of users, and both the mirror sites and users are equipped with caching memories. Although there already exist works studying this setup and proposing coded caching scheme to reduce transmission loads, two main problems are remained to address: 1) the optimal communication load under the uncoded placement for the first hop, denoted by $R_1$, is still unknown. 2) the previous schemes are based on Maddah-Ali and Niesen's data placement and delivery, which requires high subpacketization level. How to achieve the well tradeoff between transmission loads and subpacketization level for the hierarchical caching system is unclear. In this paper, we aim to address these two problems. We first propose a new combination structure named hierarchical placement delivery array (HPDA), which characterizes the data placement and delivery for any hierarchical caching system. Then we construct two classes of HPDAs, where the first class leads to a scheme achieving the optimal $R_1$ for some cases, and the second class requires a smaller subpacketization level at the cost of slightly increasing transmission loads.
翻译:本文认为,一个服务器与多个镜像站点连接,每个服务器与一组不同的用户连接,镜像站点和用户都配备了缓冲记忆。虽然已有研究这一设置的工作,并提出了减少传输负荷的编码缓冲办法,但有两个主要问题有待解决:(1) 以1美元表示的第一跳未编码位置下的最佳通信负荷仍然未知。(2) 先前的计划基于Maddah-Ali和Niesen的数据定位和交付,这需要高子包装水平。 如何在等级缓冲系统的传输负荷和子包装水平之间实现良好的平衡尚不明确。 在本文中,我们的目标是解决这两个问题。 我们首先提出一个新的组合结构,名为等级定位交付阵列(HPDDA),这是任何等级缓冲系统的数据定位和交付的特点。 然后,我们建造了两个HDDA的分类,其中第一类导致在有些情况下实现最高值R_1美元计划,而第二类则需要更小的子包装,在略增加的传输成本水平上需要更小的子包装。