Estimating traffic intensities at different nodes in networks via a probing stream

In future it will be increasingly important for the customers and the service providers to make measurements in the Internet to ensure the QoS required by various services. But not everyone will have access to all the information available at the various network components (e.g. at the routers) for such measurements. Thus it is important to devise new methods to estimate the traffic intensity at various nodes based on partial information available to a user or a service provider (say from a subnetwork not owned by it). The measurement tools available in the Internet, for example, the PING probe and the TRACEROUTE (available in various refinements) have been used by users to obtain the connectivity and some congestion information in the Internet. In this paper using these tools we suggest and compare different estimators to estimate aggregate traffic intensities at various nodes in the network. One can use this information for selection of routes as well as for admission control in the integrated and differentiated services architectures in the Internet and also in a connection admission controller in ATM networks. Our method works for general stationary traffic streams as well as for general network topologies. We also show the convergence of these estimators and provide simulation results