Hash tables – which map “keys ” onto “values” – are an essential building block in modern software systems. We believe a similar functionality would be equally valuable to large distributed systems. In this paper, we introduce the concept of a Content-Addressable Network (CAN) as a distributed infrastructure that provides hash table-like functionality on Internet-like scales. The CAN is scalable, fault-tolerant and completely self-organizing, and we demonstrate its scalability, robustness and low-latency properties through simulation.

Abstract — Application layer multicast (ALM) enables rapid deployment of multicast applications in the Internet. In ALM, application hosts organise themselves into an overlay topology on top of the underlying unicast network. Application data is multicast over the overlay network. In general, ALM protocols can be classified as either tree-first or mesh-first approach. In this paper, we provide a survey and simulation study for the class of tree-first ALM protocols. We investigate the efficiency of HMTP, TBCP, NICE and several variants of transformation-based protocols in terms of tree cost and delay optimisation. To the best of our knowledge, this paper provides the first head-to-head comparison of various tree-first protocols in a single simulation environment. Results show that depth-first searching technique in HMTP can achieve the lowest cost trees. On the other hand, a transformation-based technique that involves only local region nodes is able to construct low latency trees. In addition, we propose an enhancement to TBCP to improve its performance in producing low delay trees. D

Conventional delivery mechanisms for many-to-many multicast services consider either a single shared tree or a source rooted trees approach. These two mechanisms complement each other in terms of performance and protocol complexity. In this paper, we describe a design that uses multiple shared trees that would provide a balance between these two extreme approaches. Our proposal works at the application layer so as to enable instant deployment in the Internet. We discuss several alternative designs to realise the multiple shared trees model, and we have chosen a mesh-first approach where multiple trees are derived from a mesh overlay topology. The design is scalable as each participating node only needs to maintain a small volume of routing information. In addition, it is robust as the mesh topology provides alternate paths between the members. 1