Skip to Main Content
Server replica placement of content distribution networks (CDNs) addresses the problem of properly placing a set of replicas among a set of candidate sites to optimize system performance gain. Previous studies on server replica placement problem have focused on comparing the performances of different replica placement strategies. However, the growth of the optimal performance gain with the increase of server replicas has rarely been studied. Moreover no research has been done to study the effects of client's content demand and distribution patterns on the growth of performance gain. In this paper, these problems have been addressed in detail. The server replica placement problem is modeled as the well-known Minimum p-median problem and solved by a heuristic algorithm. A trace-driven simulation is performed and the relationship between the performance gain growth and the replica size is established A parametric study is performed to study the effect of content demand pattern α on the performance gain growth. A simple scheme of narrowing down the candidate sites is proposed to speed up the computation.
Date of Conference: 27-31 Jan. 2003