Load Sharing in Distributed VoD (Video on Demand) Systems

| In recent years, there has been an increasing interest in Video on Demand (VoD) systems. A promising way to contain the costs in such systems, is to link up several small VoD servers to a (limited capacity) network and allowing servers with idle retrieval bandwidth to help out servers that are temporarily overloaded; the goal is to minimize the waiting time for service to begin. In a previous work, some authors have proposed an algorithm called GWQ (Global Wait Queue) for exploiting the load sharing in a distributed VoD system. However they assume that the videos are replicated in all the servers and all the videos are evenly required. In our research, we introduce two main di erences: the videos are only partially replicated and they are required according to its popularity. We present an algorithm to e ciently share the load in such system and an analytical model that captures the performance of this algorithm, which we validate through simulations. In addition, we study the in uence of the replication and the popularity of the videos, in the users wait times. Keywords| Video-on-Demand, distributed system, load sharing, performance modeling, partial replication, popularity.