No abstract

The history of the Internet is that of rapid change and adaptability. The simple IP service model and the informal standardization process have fostered rapid evolution of systems and services. However, as the Internet has transformed from an academic playground to an essential commercial infrastructure the service model and the standardization process have become more complex. While outsourcing and optimizations are driving services into the network, the IETF is increasingly the battleground of vendors. Consequently, introducing new services to the Internet has become increasingly complicated.Active and programmable networking is an effort to reclaim the flexibility of the early Internet and reduce the need for standardization.In this paper we introduce a new abstraction, packet processors, to evolve the lower layer router facilities and to extend the programmable interface in a type safe manner. We show how this abstraction is sufficient to capture all facilities commonly found in data-paths of commodity routers and how it gives the ability to introduce new types of packet processors. Packet processors can add type specific methods thereby extending the programmable interface while maintaining the semantic integrity of the Node OS and the Execution Environment. We discuss how paths can be constructed from a sequence of packet processors, and how these paths can cross the serviceshode OS boundary, and cross hardware processor boundaries. Our measurements show that the added flexibility adds negligible overhead.

Abstract. For multicast to be a viable as a dependable network service, mechanisms for monitoring and managing multicast flows must be developed. Significant prior work exists to use receiver observations to derive properties of multicast groups, including the underlying distribution topology and membership size. However, just as multicast is delegation of replication and distribution from the sender into the network, we contend that similar delegation of monitoring and management into the network is warranted and beneficial. gTrace is a set of mechanisms to monitor multicast distribution trees that operates on routers participating in the multicast distribution to obtain accurate observations throughout the tree in a scalable manner. In this paper we present the protocol and validate its benefits through analysis, simulation and experimentation.

No abstract

Whether driven by security concerns, need for flexibility, deployment of advanced services or as a simplified outsourcing model, overlaying a virtual service topology over the underlying network infrastructure is common. To ensure and enforce consistent service quality, fairness and protocol behavior it is necessary to measure and monitor these service level topologies. In this paper we present extensible general purpose mechanisms to monitor and measure characteristics of a service level topology at the nodes of the topology. The mechanisms provide means to dynamically deploy a distributed observation function at the nodes of the topology and to collate the observations into a result given to the requestor on a subscription channel. These are control plane mechanisms, outside of the router datapath, where we assume programmability and low cost memory. We give several examples of how to use these mechanisms to compute interesting properties of the topology.