Fault Isolation in Multicast Trees Anoop Reddy, Ramesh Govindan, Deborah Estrin. USC/Information Sciences Institute Presented by: Srikrishna Gurugubelli
Motivation Multicasting is used for streaming video and audio. Any route change or packet loss will affect the quality. Need a method to efficiently track the fault and correct it.
Introduction Fault isolation in the context of large multicast distribution trees. Focus is on single source trees. Problem: locating on-tree router or link which is the origin of a fault. Fault: link with significant packet loss or origin of a route change resulting in the change of path from source to receiver.
Methodology of Fault Isolation A session watcher at each receiver Session watcher sends periodic probes Keeps a history of the probe responses Coordinates with other session watchers to minimize overhead and probe load.
Naïve Technique Each session watcher (w a ) periodically traces its path to the source(using mtrace) and maintains history of these traces. Using this information it can identify the link responsible for significant loss. Locating router responsible for a route change is not possible without querying other session watchers on the path. Most links are monitored by more than one session watcher Scaling problem: closer the source of distribution tree, greater the overhead of mtrace requests.
Subcast-Based Fault Isolation Relies on two capabilities –Ability of a session watcher W a to specify a hop limit h on an mtrace request. The request terminates at a router R which is h hops away from the receiver. This turnaround router generates the mtrace response. –Subcast: router support to allow a sender to specify that a packet be multicast to the subtree rooted at the specified router.
Overview If path to source overlaps, only one of the session watchers needs to monitor the overlapped segment. A session watcher W a which is at a longer distance from the common ancestor goes beyond it. The other session watcher W b will terminate its request at the common ancestor. The identity of the common ancestor is obtained by analyzing the history of responses.
Overview (contd..) If W a fails, W b will not get mtrace responses from W a. When this happens W b will extend its mtrace request all the way to the source. If W a returns to the tree, both the session watchers will monitor some links for a short time. When W b sees the mtrace responses from W a, it backs off.
Other Methods Fault Isolation Using Directed Multicast Fault Isolation Using Scoped Multicast Fault Isolation Based on Limited Multicast
Evaluation Measurements made on full n-ary trees –Maximum overhead –Average overhead –Expected error Evaluation for irregular trees –Study is made using sample trees from a random tree generator –Performance averaged across several samples
Conclusions Explored and gave a receiver-driven path probing primitive. Explored router-assisted schemes for selective forwarding of probe responses Schemes using router-assist outperform other alternatives.