Cache Placement Methods Based on Client Demand Clustering

One of the principal motivations for content delivery networks, such as those that have been deployed over the past thr ee years, is to improve performance from the perspective of the client . Ideally, this is achieved by placing caches close to groups of clients and the n routing client requests to the nearest cache. In the first part of this paper we present a new method for identifying regions of client demand. Our met hod uses best path information from Border Gateway Protocol (BGP) routin g tables to create a hierarchical clustering of autonomous systems (AS ’s). The method iteratively adds small clusters to larger clusters based onminimizing the Hamming distance between the neighbor sets of the clusters. This method results in a forest of AS trees where we define each tree root asn Internet backbone node. This forest representation of AS connectivi ty is an idealization of the Internet’s true structure. We test for fidelit y by comparing AS hop distances to the Internet backbone. One of the strengt hs of our AS clustering method is that it naturally lends itself to thecache placement problem. In the second part of this paper, we present two cach e placement algorithms based on a tree graph of demand. The algorithms ad dress the problems of placing single caches and multiple caches so as t o minimize inter-AS traffic and client response time. We evaluate the ef fectiveness of our cache placement algorithms using Web server logs and sho w that they can greatly improve performance over random cache placemen t. Keywords—Cache Placement, Hierarchical Clustering, Demand Analysis