Replication techniques for scalable content distribution in the Internet

In the traditional model of content distribution, a client contacts a server requesting the desired content and the server responds by sending the content to the client. The main problem with this model is that a server can become overloaded by re quests from a large number of clients, which can also cause congestion on the links leading to the server. Server replication, multicast content delivery and the emerging peer-to-peer paradigm are all techniques that address this problem and provide for scalable content distribution. Our research has focused on the design and evaluation of protocols for scalable content distribution in the Internet revolving around these techniques.; Replicating the servers provides clients with more locations to choose from when they desire to download content, while multicast communication allows a server to communicate with a group of receivers efficiently. These two approaches to scaling content distribution can be combined by replicating multicast servers. We describe and evaluate three protocols to perform multicast server selection that vary in their architectures and the extent of network support required for their operation.; When multicast communication is used, the scope of the multicast is also a concern. Currently, the propagation of multicast packets is controlled by two scoping methods: TTL scoping and administrative scoping. Both of these approaches require the source to control the scope of the multicast. We present multicast receiver-based scoping as a mechanism to allow receivers to place conditions on their multicast join. These conditions must be met in order for a receiver to join successfully and remain a member of the multicast group.; Multicast communication and server replication are two approaches to improve the scalability of the traditional client-server model of content distribution. However, the emerging peer-to-peer file sharing systems have a different model. With these systems, files are shared at peers that can act as both clients and servers. The current designs of these systems are lacking a scalable searching mechanism. To address this problem, we investigate the use of replication to improve the success rate of searches and reduce the searching traffic in distributed peer-to-peer file sharing systems.