Adaptive Compression Trie Based Bloom Filter: Request Filter for NDN Content Store

Abstract

In named data networking (NDN), content store (CS) is proposed to provide on-path cache service. When user’s request with content name is forwarded to NDN node, exact match in CS is carried out first. Under moderate cache hit ratio, most requests result in mismatch in CS searching process, which causes large overhead to the packet forwarding, and the overhead would rise as the scale of CS increases. In this paper, request filter of CS is studied and Compression Trie-based Bloom Filter (CT-BF) is proposed. CT-BF takes advantage of on-chip Bloom Filter to quickly filter out mismatch requests, and Compression Trie is adopted to accommodate the large CS name Trie into space-limited on-chip Bloom Filter. Optimal Compression Trie under space constraint is discussed for the first time and a heuristic approach Adapted Compression Trie based Bloom Filter (ACT-BF) is proposed for on-line operation. Simulation results show that ACT-BF can efficiently filter out mismatch requests with given on-chip space constraint and hence reduce average CS search delay.