COMP 431 The Network Layer: Routing & Addressing Internet - PowerPoint PPT Presentation

COMP 431 The Network Layer: Routing & Addressing Internet Services & Protocols Outline application ◆ Network layer services transport network network network network link network network ◆ Routing algorithms link physical link Distance Vector & Hierarchical physical physical » Least cost path computation network network algorithms link Routing physical ◆ Hierarchical routing » Connecting networks of networks network network link ◆ IP Internet Protocol physical » Addressing network network Jasleen Kaur » IPv6 link physical ◆ Routing on the Internet » Intra-domain routing » Inter-domain routing April 2, 2020 application transport network network link physical 1 2

Routing Algorithms Decentralized Routing Algorithms Taxonomy Distance Vector Routing 5 5 ◆ Iterative: ◆ Global or decentralized 3 3 C C B B » Nodes exchange cost information 5 5 information? 2 2 until each node has the current A A F F 2 route costs 2 1 1 3 3 » The algorithm is self-terminating 1 1 2 2 ◆ Global — all routers D E D E — there’s no explicit stopping point 1 1 maintain the complete graph of the network (topology, costs) ◆ Asynchronous: » “ Link state ” algorithms » Nodes need not exchange information and iterate in lock step » Intermediate results may be inconsistent across nodes ◆ Decentralized — router knows link costs to physically connected ◆ Distributed: adjacent nodes » Each node communicates only with directly-attached adjacent nodes » Run iterative algorithm to exchange information with adjacent nodes » (But there is no flooding of cost information) » “ Distance vector ” algorithms 3 4

Distance Vector Routing Distance Vector Routing Cost to Distance table data structure Distance table example Destination via Cost to Destination via D E () A B D ◆ Each node has its own table with a... D E () A B D 1 C B » Row for each possible destination 7 A 1 14 5 A » Column for each directly-attached adjacent 8 2 A 1 14 5 node (neighbor) 1 B 7 8 5 E D 2 B 7 8 5 ◆ Each table entry gives cost to reach C 6 9 4 destination via that adjacent node C 6 D E ( C , D ) = c( E , D ) + min w { D D ( C,w )} 9 4 » Distance = Cost D 4 11 2 = 2 + 2 = 4 D 4 11 2 D E ( A , D ) = c( E , D ) + min w { D D ( A,w )} = 2 + 3 = 5 A loop?! D X ( Y , Z ) = distance from X to Y via Z as first hop = c ( X , Z ) + min w { D Z ( Y , w )} D E ( A , B ) = c( E , B ) + min w { D B ( A,w )} w = { neighbors of Z } = 8 + 6 = 14 Loop! 5 6

Distance Vector Routing Distance Vector Routing Distance table example Algorithm ◆ Iterative, asynchronous: Distance Table Routing Table Each node: 1 each local iteration caused by: C B 7 D E () D E () A B D » Local link cost change, or wait for change in local A 8 2 » Message from adjacent node that its link cost or message from 1 least cost path to some destination has A 1 14 A A , 1 5 adjacent node E D changed 2 ◆ Distributed: B 7 8 B 5 D , 5 recompute distance table » Each node notifies adjacent nodes only when its least cost path to some D , 4 C 6 9 C 4 ◆ The distance table destination changes if least cost path to any gives the routing » Adjacent nodes then notify their destination has changed, D , 2 D 4 11 D adjacent nodes if this update changes a table 2 notify adjacent nodes least cost path » Just take the minimum cost per destination Adjacent Node to Use, Cost 7 8

Distance Vector Routing Distance Vector Routing Algorithm Algorithm main loop (at node X ) loop wait until ( receive link cost change to adjacent node v ◆ Initialization phase: At all nodes X : or receive new_val == min w D v (Y,w)) from v) if (c(X,v) changes by d) { /* d could be + or - */ /* change cost to all destinations via v by d */ for all adjacent nodes V { for all destinations y /* includes v */ D X ( * ,V) = “ ∞ ” /* the cost to reach all destinations through D X (y,v) = D X (y,v) + d } any neighbor is infinite */ else { if (received new_val for y from v) /* the “ * ” operator means “for all rows” */ /* shortest path from v to some y has changed */ D X (V,V) = c(X,V) /* record the cost to reach each adjacent node /* change the distance to y through v */ } (cost from X to each V) */ D X (y,v) = c(X,v) + new_val } for all destinations Y & adjacent nodes V { for all destinations y { send min w D X (Y,w) to V /* send current minimum costs for all find min_cost(y) = min w D X (y,w) /* w is all X's neighbors */ destinations to all neighbors */ if ( new min_cost (y)) { /* new minimum cost to y found */ } /* w taken over all X's neighbors */ for all adjacent nodes v send new_val = min_cost(y) to v } } forever 9 10

Distance Vector Algorithm Distance Vector Algorithm Example Example Cost via Cost via Cost via Cost via Cost via D X Y Z D X Y Z D X Y Z D X Y Z D X Y Z Dest. Y 2 ∞ Dest. Y 2 8 Dest. Y 2 ∞ Dest. Y 2 8 Dest. Y Y Y Z ∞ 7 Z 3 7 Z ∞ 7 Z 3 7 Z 2 2 Cost via Cost via Cost via Cost via D Y X Z D Y X Z D Y X Z D Y X Z 1 1 X X Dest. X 2 ∞ Dest. X 2 ∞ Dest. X 2 8 Dest. X D X ( Y , Z ) = c ( X , Z ) + min w { D Z ( Y , w )} Z ∞ 1 Z ∞ 1 Z 9 1 Z 7 7 = 7 + 1 = 8 Z Cost via Z Cost via Cost via Cost via D Z X Y D X ( Z , Y ) = c ( X , Y ) + min w { D Y ( Z , w )} D Z X Y D Z X Y D Z X Y = 2 + 1 = 3 Dest. X 7 ∞ Dest. X 7 ∞ Dest. X 7 3 Dest. X Y ∞ 1 Y ∞ 1 Y 9 1 Y Time Time 11 12

Distance Vector Algorithm Distance Vector Algorithm Example Link cost changes 1 Y 4 Cost via Cost via Cost via ◆ When a node detects a local link cost change: D X Y Z D X Y Z D X Y Z X 1 » The nodes updates its distance table » If the least cost path changes, the node notifies its 50 Dest. Y 2 ∞ Dest. Y 2 8 Dest. Y 2 8 Y Z neighbors Z ∞ 7 Z 3 7 Z 3 7 2 Cost via Cost via Cost via via via via via D Y X Z D Y X Z D Y X Z 1 D Y X Z D Y X Z D Y X Z D Y X Z X Dest. X 2 ∞ Dest. X 2 8 Dest. X 2 4 “ Good X 4 6 X 1 6 X 1 6 X 1 3 news Z ∞ 1 Z 9 1 Z 5 1 7 travels via via via via fast ” Z Cost via Cost via Cost via D Z X Y D Z X Y D Z X Y D Z X Y D Z X Y D Z X Y D Z X Y X 50 2 X 50 5 X 50 5 X 50 2 Dest. X 7 ∞ Dest. X 7 3 Dest. X 7 3 Y ∞ 1 Y 9 1 Y 9 1 c ( X , Y ) t 0 t 1 t 2 Algorithm t 3 changes terminates Time 13 14

Distance Vector Algorithm The Count to Infinity Problem Routing Loop! The “ poisoned reverse ” technique Link cost changes Does it T erminate? Y 60 4 Y ◆ Good news travels fast, but… ◆ If Z routes through Y to get to X : X 1 4 ◆ “ Bad news ” travels slow! » Then Z tells Y that Z ’ s distance to X is infinite X 50 1 Z » The “ count to infinity ” problem 50 Z Initialization… via via via via via via via via D Y X Z D Y X Z D Y X Z D Y X Z D Y X Z D Y X Z D Y X Z D Y X Z X 4 6 X 60 6 X 60 6 X 60 8 X 60 8 X 4 ∞ X 4 51 X 4 ∞ via via via via via via via via D Z X Y D Z X Y D Z X Y D Z X Y D Z X Y D Z X Y D Z X Y D Z X Y X 50 7 X 50 9 X 50 5 X 50 5 X 50 7 X 50 ∞ X 50 5 X 50 5 c ( X , Y ) t 0 t 1 t 2 t 3 Algorithm t 4 t -3 t -2 t -1 changes continues on… 15 16

The Count to Infinity Problem Least Cost Path Computations The “ poisoned reverse ” technique Comparison of the link-state & distance vector algorithms 60 Y 4 ◆ Message complexity: ◆ Robustness: what happens if ◆ If Z routes through Y to get to X : X 1 there are failures? » LS: With N nodes, E links, O ( N x E ) » Then Z tells Y that Z ’ s distance to X is infinite 50 messages sent for flooding » LS: Node can advertise incorrect Z link cost ◆ (Will this completely solve the problem?) » DV: Exchange between neighbors only (may trigger further Each node computes only its own exchanges) table via via via via via » DV: Node can advertise incorrect D Y X Z D Y X Z D Y X Z D Y X Z D Y X Z ◆ Speed of Convergence: path cost » LS: O (N 2 ) algorithm and O ( N x E ) Each node’s table used by others X 4 ∞ X 60 ∞ X 60 ∞ X 60 51 X 60 51 messages ❖ Errors propagate through network ❖ May have oscillations » DV: Convergence time varies via via via via via D Z X Y D Z X Y ❖ Routing loops possible D Z X Y D Z X Y D Z X Y ❖ Count-to-infinity problem X 50 61 X 50 ∞ X 50 5 X 50 5 X 50 61 c ( X , Y ) t 0 t 1 t 2 t 3 Algorithm t 4 changes terminates 17 18