#1 EETS 8316/NTU CC725-N/TC/ Routing - Circuit Switching  Telephone switching was hierarchical with only one route possible —Added redundant routes —Evolved to full mesh Subscribers Central offices Toll switches

#2 EETS 8316/NTU CC725-N/TC/ Routing - Packet Switching  Data networks have used various routing protocols —Flooding —Source routing —Distributed shortest-path routing Distance-vector routing Link-state routing  Layered routing is used in Internet for scalability

#3 EETS 8316/NTU CC725-N/TC/ Flooding  Packet is copied at every router and broadcast to neighbors —Useful when info must be shared at all routers, eg, routing updates  Advantage: no knowledge of network topology is necessary —Well suited to highly dynamic networks, eg, battlefield  Disadvantage: Number of message copies may be huge without additional mechanisms, eg:

#4 EETS 8316/NTU CC725-N/TC/ Flooding (Cont) —Messages have hop limit to restrict number of copies —Messages are uniquely identified so routers can stop unnecessary copies —Routers can establish minimum spanning tree that message copies will follow Overlay tree that covers every router exactly once Example spanning tree Example network

#5 EETS 8316/NTU CC725-N/TC/ Source Routing  Sender determines route and specifies it in packet header —Example: IP source routing option —Usually centralized route server maintains network topology info and collect updates from all routers  Advantage: sender can select optimal route or test certain routes  Disadvantages: —Centralized routing is not scalable to large networks —Difficult to carry long routes in packet header

#6 EETS 8316/NTU CC725-N/TC/ Distributed Shortest-Path Routing  Distributed shortest-path routing: each router builds up routing table independently and shares routing updates with other routers —Routing table entries specify how to forward packets to destination addresses/networks  Two main types of routing protocols known from Internet experience: —Distance-vector or Bellman-Ford —Link-state

#7 EETS 8316/NTU CC725-N/TC/ Distance-Vector Routing (Bellman-Ford)  Example: RIP (routing information protocol)  Each router maintains a table (vector) of reachable destinations and corresponding shortest distances to each destination —Periodically shares routing table with its neighbors  Each router modifies its routing table with the received updates —Shortest distance to every destination is chosen —New reachable destinations are added, unreachable destinations are deleted

#8 EETS 8316/NTU CC725-N/TC/ Distance-Vector Routing (Cont)  RIP was first routing protocol used in Internet but difficulties were found —Slow to propagate routing updates through large network —Routers can have inconsistent or even incorrect information —Scalability problem: vector of destinations becomes longer as network is larger  routing updates are larger

#9 EETS 8316/NTU CC725-N/TC/ Link-State Routing  Examples: OSPF (open shortest path first), IS- IS (intermediate system - intermediate system)  Each router maintains a graph representing complete network topology —Monitors state of its links with neighboring routers —Floods link-state updates to all routers periodically  Each router receives link-state updates from other routers and revises its graph —Shortest routes can be computed systematically by Dijkstra’s algorithm

#10 EETS 8316/NTU CC725-N/TC/ Link-State Routing (Cont)  Example: topology graph is given, minimum spanning tree generated by Dijkstra’s algorithm is all shortest routes from source node A (link labels are distances) A A

#11 EETS 8316/NTU CC725-N/TC/ Link-State Routing (Cont)  Advantages: —Every router should have consistent, complete topology info  consistent, optimal route selection —Routing updates do not grow with size of network (link-state updates depend on number of links per router)  Disadvantage: flooding link-state updates can result in large traffic volume and traffic oscillations —Limit updates to only significant changes —Flooding is limited by layered routing

#12 EETS 8316/NTU CC725-N/TC/ Layered Routing  Internet is too large for routing as “flat network” where every router is aware of every other router  Layered routing: routers are grouped together into “routing domains” or “areas” —At lowest layer, interior routers within each routing domain use an intra-domain routing protocol with each other —Exterior routers represent each routing domain in higher layer inter-domain routing protocol —Interior routers are aware only of routing domain

#13 EETS 8316/NTU CC725-N/TC/ Layered Routing (Cont) Interior routers use intra-domain routing Routing domain Exterior routers use inter-domain routing

#14 EETS 8316/NTU CC725-N/TC/ Ad Hoc Networks  Wireless networks may be classified as —Infrastructured: cellular systems consist of fixed (wireline) network with mobile users attached to network edge through fixed base stations Handoff is necessary and critical —Ad hoc: no fixed infrastructure All nodes can be mobile and autonomous Connections are dynamic and arbitrary Examples: battlefield,, emergency rescue operations, WLAN, PAN

#15 EETS 8316/NTU CC725-N/TC/ Ad Hoc Networks (Cont)  IETF (Internet Engineering Task Force) Mobile Ad Hoc Networks (MANET) working group is studying proposals —MANETs consist of mobile routers and mobile hosts (can be physically same) —Mobile hosts are temporarily or permanently attached to a mobile router —Inter-router connectivity can change frequently Routing protocol must adapt to continually changing topology Addresses cannot be based on location

#16 EETS 8316/NTU CC725-N/TC/ Ad Hoc Networks - Routing  Routing protocols —Table driven: nodes maintain and update routing tables Cost in overhead traffic for distributing routing updates Routes are always ready when needed for data —On-demand (source initiated): routing tables are not maintained cont., routes are found only when needed for data transmission Cost in overhead traffic only when data is ready for transmission Routing tables are smaller

#17 EETS 8316/NTU CC725-N/TC/ Ad Hoc Networks - Table Driven Routing  Destination sequenced distance-vector (DSDV) routing —As in Bellman-Ford, each node maintains routing table of reachable destinations and their distances (number of hops) —Routing table info. is periodically broadcast for consistency, using two packet types: Full dump packets contain complete routing info. sent infrequently Incremental packets contain only new routing info.

#18 EETS 8316/NTU CC725-N/TC/ Ad Hoc Networks - Table Driven Routing  Clusterhead gateway switch routing —Nodes are grouped into clusters and elect a cluster head by a cluster head selection algorithm —DSDV routing is used within clusters —Between clusters, packets are routed to cluster head  gateway  cluster head Gateways are nodes within communications range of 2 or more cluster heads —Packets may have to hop through multiple cluster heads and gateways to final destination

#19 EETS 8316/NTU CC725-N/TC/ Ad Hoc Networks - Table Driven Routing —Each node maintains “cluster member table” of cluster heads for every destination node Nodes broadcast their cluster member table and update their own tables Packet Cluster head Cluster head Gateway Cluster head

#20 EETS 8316/NTU CC725-N/TC/ Ad Hoc Networks - Table Driven Routing  Fisheye link-state routing —Similar to link-state routing but link-state updates are not broadcast to every node Broadcast link-state updates to nodes one hop away every 1 sec, nodes 2 hops away every 10 sec, nodes 3 hops away every 50 sec,... —Routes are more accurate to nearby nodes, but approximate to distant nodes Packets will converge to more accurate routes to destination as get closer to destination

#21 EETS 8316/NTU CC725-N/TC/ Ad Hoc Networks - Table Driven Routing Link state updates 1 sec Link state updates 10 sec Link state updates 50 sec

#22 EETS 8316/NTU CC725-N/TC/ Ad Hoc Networks - Source Initiated Routing  Ad hoc on-demand distance-vector routing —Similar to DSDV routing —Source node with data ready to send initiates a path discovery process Broadcasts route request (RREQ0 packet to its neighbors. they broadcast to their neighbors, and so on Eventually, destination node responds with route reply (RREP) packet in reverse direction Nodes build up routing tables

#23 EETS 8316/NTU CC725-N/TC/ Ad Hoc Networks - Source Initiated Routing RREQ S D RREP S D

#24 EETS 8316/NTU CC725-N/TC/ Ad Hoc Networks - Source Initiated Routing  Dynamic source routing —Nodes maintain route caches —If route to destination is unknown in its route cache, node will initiate route discovery Broadcasts route request packet with destination address and source address Other nodes will add their address into packet and continue broadcast to other nodes Route reply is returned by destination or a node that knows a route to destination

#25 EETS 8316/NTU CC725-N/TC/ Ad Hoc Networks - Source Initiated Routing  Associativity-based routing —Routes are selected base on new metric: degree of association stability Each node periodically generates a beacon to advertise presence Neighboring nodes will increment associativity tick of this node in their associativity tables for each beacon received Association stability is defined by connection stability between two nodes over space and time –High association stability may mean low node mobility

#26 EETS 8316/NTU CC725-N/TC/ Ad Hoc Networks - Source Initiated Routing —Route discovery: Node will broadcast query (BQ message) to destination Intermediate nodes will add their address and associativity ticks into message Destination node receives packets with associativity ticks of nodes along routes taken, and chooses route with best degree of association stability Destination node returns BQ-REPLY back to source along chosen route Intermediate nodes will record this chosen route