Computer Networks II

Transcription

1 ipartimento di Informatica e Sistemistica omputer Networks II Routing protocols Overview Luca Becchetti Luca.Becchetti@dis.uniroma.it /200

2 Goals escribe approaches and give overview of mechanisms adopted escribe key aspects of implementations nalyze two solutions adopted in practice RIP OSPF (more in detail) Some aspects to addressed in more detail by student [see references]

3 Routing Routing protocol Goal: find good sourcedestination path Graph abstraction: Nodes are the routersr Links represent subnets Weights on links: delay, cost in, level of congestio Special case: P2P link 2 5 B Good route: E Usually, shortest path different definitions of shortest path possible 5 2 F

4 lassification of routing algorithms Global vs distributed Info Global: ll nodes have full info about graph and link costs Link state algorithms istributed: Router knows neighbours and costs of links to them Iterative and distributed computation: info exchanged with neighbours istance vector algorithms Static vs dynamic Static: Routes vary slowly over time ynamic: Routes change frequently Periodic updates Updates triggered by changes in links' costs

5 Forwarding (or routing) table ost to reach a given destination Next hop (router) to destination This is only neccessary info Real tables much more complex 4 B 3 9 E 6 2 F estinazione E F E,2,4 Outgoing link, cost E,2 E, E,3 Routing table at B

6 Overview Forwarding vs Routing forwarding: to select an output port based on destination address and routing table routing: process by which routing table is built Factors static: topology dynamic: load Summarizing: Routing: find routes to possible destination, build Routing Table Forwarding: forward incoming packet to neighbour router on the basis of packet's destination, using information stored in routing table 4 B 3 9 E 6 2 F

7 istance Vector lgorithms overview Base algorithm: Bellman Ford (distributed version) Each node maintains a set of triples (estination, ost, NextHop) irectly connected neighbors exchange updates periodically (on the order of several seconds) whenever table changes (called triggered update) Each update is a list of pairs: (estination, ost) If receives (, 0) from B (B has to be a neighbour) this means that cost to reach over is 0 + cost(, B) Update local table if receive a better route smaller cost came from next hop Refresh existing routes; delete them if they time out

8 Evolution at node ynamic evolution B 3 2 Initial table at est ost NextHop B B 4 inf - 4 Update from B est ost 2 3 entry for missing fter update from B est ost NextHop B B 3 B 4 B Update from est ost B 2 entry for missing

9 istance Vector Routing lgorithm Iterative: Iterates until no more info exchanged self terminating: no stop message exists synchronous: Nodes do not perform steps synchronously istributed: Every node communibates only with immediate neighbours istance Table (distance table present in every node) One entry for each destination One column for every node's neighbour Example: at node X, for dest. Y through Z: X (Y,Z) = = istance from X to Y through Z Z c(x,z) + min { (Y,w)} w

10 istance table: example 7 E (,) E (,) E (,B) 5 B E 8 2 = c(e,) + min { (,w)} w = 2+2 = 4 = c(e,) + min { (,w)} w = 2+3 = 5 iclo! 2 B c(e,b) + min { (,w)} w = = 8+6 = 4 iclo! osto per la destinazione via E () estinazione B B

11 istance table gives RT E () ost to destination via B Outgoing link, costo 4 5, estinazione B estinazione B,5,4,2 istance table Routing Table

12 istance Vector Routing: overview Iterative, asynchronous: each local iteration caused by: local link cost change message from neighbor: its least cost path change from neighbor istributed: each node notifies neighbors only when its least cost path to any destination changes neighbors then notify their neighbors if necessary Each node: wait for (change in local link cost or msg from neighbor) recompute distance table if least cost path to any dest has changed, notify neighbors

13 istance Vector lgorithm t every node X: Initialization: 2 for every adjacent node v: 3 X(*,v) = infinity /* * means for every entry" */ X 4 (v,v) = c(x,v) 5 for every destination y X 6 send min (y,w) to every neigh. /* w varies over all neighbours of X */ w

14 istance Vector lgorithm (cont.): 8 loop 9 wait (until link cost to some neighbour V changes 0 or update from neighbour V received) 2 if (c(x,v)changes by an amount d) 3 /* update link costs for all destination over X by d 4 /* note: d may be positive or negative */ X X 5 for every destination y: (y,v) = (y,v) + d 6 7 else if (update from V for destination Y) 8 /* shortest path from V to some Y changed */ v 9 /* V sends new valu for min w (Y,w) */ 20 /* let "newval" be new value */ X 2 for every y: (Y,V) = c(x,v) + newval 22 X 23 if new value for min w(y,w) towards destination Y X 24 send new value for min w (Y,w) to every neigh forever

15 Example Table at B B estination ost NextHop Initial E F G F E Full example on Peterson and avie's book G estination ost NextHop Final 2 E 2 F 2 G 3

16 Routing Loops Unit costs. Link (, B) faults, cost from to B > inf. B Routing Table at est ost NextHop B B 2 n routers between and B receives periodic update from before sending own update updates entry for B (entry becomes (B 3 )) and sends update to Process iterates until cost from to B through becomes lower than cost through

17 Loop Breaking Heuristics lgorithm converges but this can be slow Fixes Set infinity to 6 Split horizon Split horizon with poison reverse o not send update to a neighbour if update caused by that neighbour's previous updates Work often but not always

18 Link state algorithms Base algorithm: ijkstra Every router needs to know network topology (connectivity and link costs) Implementation aspects Information gathering Flooding efine cost metrics link state protocol: OSPF (Open Shortest Path First) ijkstra's algorithm Protocol for the exchange of link state information dvanced aspects Hierarchical routing Scalability > route summarization

19 ijkstra's algorithm Every nodes knows topology and link costs an be achieved using link state broadcast ll nodes posses same info Every node computes shortest paths to all other nodes i Provides routing table for node Iterative: after k iterations, shortest paths to k destinations found Notation c(i,j): link cost between i and j. (v): last estimated value for shortest path cost to destination V p(v): v's predecessor on shortest path to v N: set of nodes for which shortest path cost has been exactly computed

20 ijkstra's algorithm/cont. Initialization: 2 N = {} 3 for all nodes v 4 if v adjacent to 5 then (v) = c(,v) 6 else (v) = infty 7 8 loop 9 find w not in N such that (w) is a minimum 0 add w to N update (v) for all v adjacent to w and not in N: 2 (v) = min( (v), (w) + c(w,v) ) 3 /* new cost to v is either old cost to v or known 4 shortest path cost to w plus cost from w to v */ 5 until all nodes in N

21 ijkstra's algorithm: esempio Step start N E EB EB EBF (B),p(B) 2, 2, 2, (),p() 5, 4, 3,E 3,E (),p(), (E),p(E) infinity 2, (F),p(F) infinity infinity 4,E 4,E 4,E Routing Table at the end 2 5 B E 5 2 F Note: every sub-path of a shortest path is a shortest path between its end points

22 Properties omputational complexity: n nodi Every iteration: check all nodes w not in N n*(n+)/2 comparison worst case: O(n^2) More efficient heap based implementation: O(nlogn) Route oscillations possiblei: e.g., link cost = congestion Why this metric? +e 0 0 B 0 e e Start 2+e 0 B 0 +e 0 recompute routing 0 2+e 0 B 0 +e recompute 2+e 0 +e B 0 e recompute

23 Implementation Strategies for link state information collection

24 Link State ijkstra requires full knowledge of network at every router Technique: Reliable flooding Node sends own link state info to all neighbours node receiving link state info forwards it to neighbours Periodic updates Use TTL to eliminate obsolete information

25 Link State Packets I of node that generated packet List (neighbour, cost of link to neighbour) SEQNO TTL LSP generation Timer expires Topology/link costs change

26 Link State Packets Reliable flooding store most recent LSP from each node forward LSP to all nodes except one that sent it generate new LSP periodically increment SEQNO start SEQNO at 0 when reboot decrement TTL of each stored LSP discard when TTL=0 X X Example: LSP originating from X and received by B (X, ((, 0), (, 2)), 5, 5) B (a) X B B (b) X B I Neigh. ost SEQNO TTL (c) (d)

27 Link State recap. For every router at start up or by any change : send LS (Link State dvertisement) LS contains LSP with link state info for all links connected to router Router exchange link state info using flooding s soon as router's B complete: compute Shortest Path Tree towards all other routers Generate routing table: Generic entry: (dest, cost, next hop) Update upon changes

28 References TP/IP guide: lso: J. F. Kurose and K. W. Ross. omputer Networking: Top own pproach, 4/E, hapter 4 (4.5 and 4.6)