Coded Caching with Low Per-Request Complexity

Abstract

The massive growth in high-rate streams (mostly video) over the Internet warranted the establishment of content distribution networks (CDNs), which bring the content closer to the consumer and hence reduce delays and congestion in the network. In this paper we address the setup of multiple Internet service providers (ISPs) sharing a federated cache. When multiple ISPs with different demands share a cache, storage cost can be reduced when the cache is coded. We address the problem of constructing the coded cache to fulfill all demands from all ISPs, where the constructions simultaneously optimize the cache's three main operational costs: storage, communication, and computation (to encode and decode requested objects). When only the storage needs to be minimized, the code-design problem can be formulated as an index coding problem, and solved by known tools in network coding. However, in practical setups the cache needs to respond efficiently (in communication and computation) to individual content-object requests, which renders known index/network-coding solutions non-applicable and non-scalable. We first show constructively that when the number of ISPs is less than or equal to 4, there is an efficient algorithm that achieves optimal storage with only bit-wise XOR operations, and with guaranteed low computation cost for serving requests. For larger numbers of ISPs, we present randomized algorithms that achieve optimal storage with XOR-only coding and low computation complexity per requested objects.