The current Internet today hosts several extensions for indirection like Mobile IP, NAT, proxies, route selection and various network overlays. At the same time, user-controlled indirection mechanisms foreseen in the Internet architecture (e.g., loose source routing) cannot be used to implement these extensions. This is a consequence of the Internet’s indirection semantics not being rich enough at some places and too rich at others. In order to achieve a more uniform handling of indirection we propose SelNet, a network architecture that is based on a virtualized link layer with explicit indirection support. Indirection in this context refers to user-controlled steering of packet flows through the network. We discuss the architectural implications of such a scheme and report on implementation progress.

The Internet has successfully promoted address uniformity and a node centric forwarding semantics. However, NAT and wireless networks among others have shown the advantage and the need of revising basic assumptions of the Internet model. In this paper we review several of these basic networking concepts and introduce a new set of network abstractions like ”membranes ” which are individual physical or virtual networks and ”wormholes ” which link one or more membranes together. This leads us to an active network architecture called SelNet that is based on tunnelling and translation mechanisms. Besides the architecture we present several network services and abstractions that can be built on top of it. A brief status report on a prototype implementation is also provided.

In this paper we study the potential and limitations of active networks in the context of adaptive applications. We present a survey of active networking research applied to adaptive applications, and a case study on a layered multicast active application. This active application is a congestion control protocol that selectively discards data in the active routers, and prunes multicast tree branches affected by persistent congestion. Our first results indicate that active networks can indeed help such an application to adapt to heterogeneous receivers, with a minimum amount of state overhead, equivalent to that of a single IP multicast group.