Optimal chunking and partial caching in information-centric networks

Abstract

Caching is widely used to reduce network traffic and improve user experience. Traditionally caches store complete objects, but video files and the recent emergence of information-centric networking have highlighted a need for understanding how partial caching could be beneficial. In partial caching, objects are divided into chunks which are cached either independently or by exploiting common properties of chunks of the same file. In this paper, we identify why partial caching is beneficial, and propose a way to quantify the benefit. We develop an optimal n-Chunking algorithm with complexity O(ns2) for an s-byte file, and compare it with ∊-optimal homogeneous chunking, where ∊ is bounded by O(n-2). Our analytical results and comparison lead to the surprising conclusion that neither sophisticated partial caching algorithm nor high complexity optimal chunking are needed in information-centric networks. Instead, simple utility-based in-network caching algorithm and low complexity homogeneous chunking are sufficient to achieve the most benefits of partial caching.