Fast Content Store Lookup Using Locality-Aware Skip List in Content-Centric Networks

Abstract

Today's Internet usage evolves rapidly from host-to-host communication to content dissemination. Content-Centric Networking (CCN) is proposed to embrace this trend by providing in-network caching capability via built-in Content Store (CS). However, to satisfy per-packet queries in high-speed networks, CS suffers a noticeable performance penalty and probably becomes the bottleneck of the entire packet forwarding system. The state-of-the-art adopts skip list as CS's underlying data structure. But, due to its O(logn) search complexity, it still has performance issues. In this work, inspired by the extensively existed temporal and spatial locality in content retrieval, we propose locality-aware skip list for performance improvement. In our design, since a packet is likely to share a prefix with previous packets, the search in the skip list could start directly from near the previously reached node of that prefix. This avoids searches repeatedly starting from the head node thus saves considerable search time. Extensive evaluation shows our design can achieve a 3× speedup over the original design on an ×86 machine.