Phonternet
This project is motivated by Prof. S. Keshav’s work on multi-mode networks.
Providing qualities of service that are customary in the phone networks on the Internet seems to be a daunting if at all feasible task. The Internet has the best effort delivery model and the telephone network provides guaranteed service. Voice traffic is constant bit rate and is well understood; data traffic on the Internet is not well understood, ever changing in character and bursty on all time scales. We believe that different networks will coexist at least in the near future. There are many interesting applications that need to use services provided by both networks. Consider, for example, an audio conferencing application which has audio and a whiteboard. For the audio we may want to use the phone network, while we want to use the Internet for the shared whiteboard. This is an example of what we call multi-mode communication. It tries to leverage the strengths of each network and provide interoperability in the control plane. In this project we develop a system to use the phone network as the data plane and the Internet as the control plane for distribution of live content.
Active Network Framework
We define active networks as networks that allow programs called agents or applets to be executed in internal network nodes such as multiplexors, firewalls, routers and switch controllers. Rudimentary active network components are already used for mobile communication (for forwarding packets after a handoff), in video conferencing (for prioritized dropping of multicast groups), and firewalls (for allowing customizable filters). The difficulty in building an active network has been the lack of a framework that exposes the interfaces available on internal nodes. In this project we build such a framework by exposing BSD nodes to the environment on each router. Reference
http://www.cs.cornell.edu/home/sharma/papers/nf.ps provides an overview of active networks.
Traffic management and monitoring for applications using WinSock 2.0
An enterprise or campus may contain many routers and machines under the control of a single administrator. In this situation, it is often a good idea to centralize some functions usually associated with individual routers. For example, a central route server can compute loop-free routes for the entire enterprise and load them into each router’s forwarding tables. A similar approach can be taken for loading multicast forwarding entries, thus freeing the routers from the burden of participating in complex multicast routing protocols. We call this ‘enterprise level management’.
The enterprise level approach also makes it easier to implement global policies. For example, an organization may want to limit the total amount of resources dedicated to multicast traffic. Although the mechanisms to restrict traffic (like policers) may be implemented at each router, the computation of parameters for the individual policers can be centralized. This project focuses on adapting Winsock to obtain statistics about usage of individual workstations and the subsequent incorporation of mechanisms to implement policies.
Framework for building reliable multicast based applications, such as white board, one way audio.
This projects focuses on building a new protocol family (SOCK_SRM) and exposing it to applications for the development of reliable multicast based applications.
Network Health Channel
This project builds on previous work on network topology discovery (
http://www.cs.cornell.edu/skeshav/papers/discovery.ps) and ping-trees to find the hot spots in the network. We use the previous work to publish information about the health of the network on an active channel.