An improved caching scheme for nonuniform demands and its optimal allocation

In content delivery networks, as more contents are being generated and the size of contents is increasing, caching has been identified as a solution to cope with congestion in communication links. The idea is to prepopulate users' cache with parts of files in off-peak time to ease the peak-time delivery traffic. In this paper, we consider nonuniform demands, where different files have different probability of being requested. We consider networks with a server connected to multiple users through a shared link. The goal is to design cache placement schemes that minimize the transmission rate from the server during peak-time content delivery. In this paper, we analytically prove that a caching scheme proposed by Zhang et al., which they showed to achieve within a constant multiplicative and additive gap from the optimal rate, is indeed optimal when there are only two users.