Skip to Main Content
In recent years, because of the exponential growth of the BGP routing table size in the Default Free Zone, the Internet has been facing serious routing scalability problems. To solve these problems, the research community in the IRTF has proposed a new Internet architecture which separates edge networks from core networks. This solution needs to map an edge address to a core address in the core networks, and cache is often used to improve its performance by storing the mapping information of some frequently used edge prefixes. In this paper we present a new cache replacement policy: Least Frequency- Delay-Product (LFDP). In LFDP, a replacement decision is made according to both the hit count of each cache entry and the latency of retrieving the corresponding mapping information, while entries used more recently or retrieved more slowly are less likely to be replaced. We use our event-driven simulation to demonstrate that LFDP is more suitable for distributed mapping systems in the core/edge separation architecture than other known cache schemes.