Ad hoc Routing: Issues and AlgorithmsMobile Ad Hoc Networking (MANET)Goals for Ad Hoc RoutingGoals for Ad Hoc Routing (contd)Some Architectural ConsiderationsBrief Review on Shortest-path RoutingDesign Options in Ad Hoc RoutingDestination-Sequenced Distance-Vector Routing (DSDV)Slide 9Temporally-Ordered Routing Algorithm (TORA)PowerPoint PresentationAd Hoc On-Demand Distance Vector (AODV)Slide 13AODVDynamic Source Routing (DSR)Slide 16Further DiscussionsNew Trends in Ad Hoc RoutingAd hoc Routing: Issues and Algorithms•Why we study this paperidentify the issues of routing support in mobile ad hoc networksunderstand some design options for ad hoc routing support schemesalgorithm overviews of a few algorithms•New trends in ad hoc routingMobile Ad Hoc Networking (MANET)•No backbone infrastructure, typically operate over a single wireless channel•dynamic topologies: nodes are free to move arbitrarily•bandwidth-constrained, variable capacity links•energy-constrained operation•limited physical security•may have both unicast and multicast/broadcast traffic •typically assume reliable broadcast at the MAC/link layer•in the near term, MANET functions as stub networks (RFC2501): all traffic carried by MANET nodes will either be sourced or sinked within the MANETNOT as transit networks carrying traffic that enters and then leaves MANETGoals for Ad Hoc Routing•Near-term goal (RFC2501):provides for effective operation over a wide range of mobile networking “contexts” (i.e. a set of MANET characteristics)supports connectionless IP service: IP layer mobile routingreacts efficiently to topological changes and traffic demands while maintaining effective routing in a MANET context•Qualitative properties of MANET routing protocols:distributed operation: no centralized solutionloop-freedom: avoid problems such as a small fraction of packets spinning around in the net for arbitrary time periodsdemand-based operation: routing algorithms adapt to the traffic pattern in the net on a demand or need basisproactive operation: flip-side of demand-based operationsecurity, sleep period operation, unidirectional link support.Goals for Ad Hoc Routing (contd)•Network contextnetwork size (# of nodes), network connectivity (avg degree of a node), topological rate of change, link capacity, fraction of unidirectional links, traffic patterns, mobility (more on topological correlation), fraction & frequency of sleeping nodes•Quantitative metrics to assess ad hoc routing protocols:end-to-end data throughput and delayroute acquisition time: of particular concern with on demand routing algorithmspercentage of out-of-order deliveryefficiency: internal measure of its effectiveness•average number of data bits transmitted/data bit delivered•average number of control bits transmitted/data bit delivered•average number of control and data packets transmitted/data packet deliveredSome Architectural Considerations•Link level operationability to detect link appearances & failures in the presence of changing topology:•periodic probing using overhead packets (faster link status sensing)•lack of link-level ACKs of transmitted message packets (slower sensing, no overhead)•shortest path routing: is it worth the cost ?changing topology: slow versus fastcongested network: light versus heavyhow to handle the gray area ? adaptive•hard state versus soft stateBrief Review on Shortest-path Routing•Internet supports shortest-path routing•Properties for shortest path:subpaths of shortest paths are shortest pathsthe weight of a shortest path is the sum of the weights of its subpaths•Dijkstra’s algorithm: builds up a shortest path tree by computing the weighted breadth first tree from the source•Bellman Ford algorithm: builds a shortest-path tree by incrementally expanding the permissible set of nodes that are allowed to be intermediate hops•Distributed Bellman Ford Algorithm•distance vector and link state routingdistance vector: exchange vectors of distance estimates to destinations (e.g. RIP)link state: maintain the entire network topology at all routers (OSPF)Design Options in Ad Hoc Routing•Who serve as routers:every mobile host is a routera subset of the nodes are routers (e.g., due to energy concerns)every host is a router but only a subset of routers are used to generate routes (virtual backbone concept)•what is the neighborhood info to propagatelink state, distance vectors, known routes•metrics for route computationweighted shortest path, interference metrics, congestion metrics, power/energy metrics, throughput metrics •how many routes are maintainedsingle route for a source-destination pairmultiple routes for a src-dest pair•when to generate routesstatic versus on demandDestination-Sequenced Distance-Vector Routing (DSDV)•Based on the Bellman Ford routing algorithm with improvements such as freedom from loops in routing table•each node maintains a routing table recording all of the possible destinations and the distance to each destinationeach entry is marked with a sequence number assigned by the destination node•routing tables are periodically transmitted throughout the network •route updates via two types of packets:full dump packets carrying all available routing infoincremental packets carrying the change only since last full dumpDestination-Sequenced Distance-Vector Routing (DSDV)•Operations using sequence numberassociate monotonically increasing sequence number with the neighborhood update information. When info about a destination is being passed, the latest sequence number generated by that destination is used. This enables source nodes to distinguish fresh info from stale infouse sequence numbers to make routing loop-free•each node generates an even sequence number every time it propagates its neighborhood update•whenever a node detects a loss of connectivity with a neighbor, it creates a neighborhood update, with an odd sequence number that is generated by incrementing the known seq number for the neighborhood by 1•if the destination generates a new neighborhood update, it receives precedence over the loss of connectivity updatereduce the frequency of updates by using some stability heuristics: damping route update propagation by an amount of time it takes for a route to settle once a change has been
View Full Document