Multicast provides scalable mechanisms to reach hundreds or thousands ofsimultaneous receivers in the network. Within the last decade, there has been asignificant effort in developing necessary protocols for multicast and integratingmulticast into the Internet as a transparent service. During this time, it hasbeen well recognized that the success of multicast highly depends on the availabilityof good monitoring and management systems. In this dissertation, wehave focused on this aspect of the problem. We have identified the characteristicsof multicast monitoring tasks and developed mechanisms and protocolsto perform a set of critical monitoring tasks.For any new service in the Internet, one of the first management tasksis to verify the availability of the service and maintain its robustness in thenetwork. This task requires an ability to detect and isolate potential problemsin the network. In the first part of our work, we have mainly focusedon developing systems/protocols to detect the existence of potential multicastproblems in the network. For multicast, problem detection requires monitoringmulticast data transfer activity in the network. This monitoring is usuallyreferred to as reachability monitoring. We have first developed an applicationlayerreachability monitoring system called sdr-monitor. This system has beenvery useful to the multicast community in detecting potential problems in thenetwork. Next, we have used this experience to develop a more systematic apxproach (a generic architecture) for multicast reachability monitoring systems.And, finally, we have used this architecture to evaluate and improve an alreadyexisting protocol, the Multicast Reachability Monitor (MRM) protocol.In the second part of our work, we focused on problem isolation for multicast.Since the multicast service depends on forwarding trees that are createddynamically by routers in the network, problem isolation usually requirescollecting multicast tree topologies from the network. In our work, we havedeveloped a mechanism, called tracetree, for discovering multicast tree topologiesfrom the network. Tracetree uses the existing multicast forwarding statefor topology discovery. Compared to alternative approaches, tracetree is moreeffective, more scalable and more efficient in discovering tree topologies in thenetwork. |