Wireless Internet Routing

Transcription

1 Wireless nternet Routing Wireless Routing in the eginning 1 Wireless nternet Routing This course: routing in T/ad hoc networks and mesh networks o nternet? = ateway(s) wireless/wired 2 1

2 Outline Review of last time Wired routing algorithms: why are they not usable in wireless scenarios n the beginning: (unicast) routing for T o Proactive vs on-demand o R, OV, V 3 Review: esh etworks with nternet ccess Wireless infrastructure (backbone) o tatic backbone ateways to the nternet o g: reifunk, eattle Wireless lients connects to P o Transparent wireless network o obility, roaming P P P nternet P nternet P 4 2

3 Review: ulti-op Wireless etworks o infrastructure: ad-hoc network o T: obile d oc etworks o ensor networks o VT Origins: packet radio networks (1978, dvances in packet radio technology ) odes: transmit, receive, forward odes can be mobile Wireless routing becomes necessary 5 Review: Wireless Physical ayer and its effect on the upper layers Unlike a wired connection! haracteristics o Time-varying channel: propagation, connectivity and available capacity o hared medium: interference, collisions onsequences for the upper layers: links exhibit o Time-varying behavior: connectivity, rate, delay o ow reliability: packets are lost (typical 10^-2 PR) o maller bandwidth than wired counterpart (10 to 100 bit/s) o symmetric / bidirectional and unidirectional links 6 3

4 Review: Time-Varying nvironment onnectivity is not a unit-disc Variable packet loss Received power is time-varying 7 Review: What is Routing ontrol Plane: Topology discovery o eighborhood discovery Route discovery and maintenance o ault recovery ata Plane: orwarding 8 4

5 hallenges of Wireless Routing Routing in a wireless network = routing in a network with: o igh time variability o ynamic topology o inks: unreliable, asymmetric, time-varying 9 etour: ow is Wired Routing Working? istance vector algorithms (e.g. RP) o ecentralized information o Router knows physically-connected neighbors, link costs to neighbors o terative process of computation, exchange of info with neighbors ink-state algorithms (e.g. OP) o lobal information o ll routers have complete topology, link cost info 10 5

6 istance Vector Routing lgorithm ecentralized algorithm: Router knows its neighbors and link costs to neighbors terative computation, exchange of info with neighbors ellman-ord quation (dynamic programming) efine d x (y) := cost of least-cost path from x to y Then d x (y) = min {c(x,v) + d v (y)} v where min is taken over all neighbors v of x alculate direction and distance to any link in a network 11 istance Vector lgorithm terative, asynchronous: ach local iteration caused by: o ocal link cost change o V update message from neighbor istributed: ach node notifies neighbors only when its istance Vector changes o eighbors then notify their neighbors if necessary o Region not concerned by topology change are not affected ach node: wait for (change in local link cost of msg from neighbor) recompute estimates if istance Vector to any dest has changed, notify neighbors 12 6

7 ink-tate Routing lgorithm et topology, link costs known to all nodes o ccomplished via link state broadcast o ll nodes have same info omputes least cost paths from one node ( source ) to all other nodes o ives routing table for that node xample: ijkstra s algorithm o terative: after k iterations, know least cost path to k dest. s otation: ijkstra s algorithm c(i,j): link cost from node i to j. cost infinite if not direct neighbors (v): current value of cost of path from source to dest. v p(v): predecessor node along path from source to v : set of nodes whose least cost path definitively known 13 ink tate Routing ach node periodically floods status of its links ach node re-broadcasts link state information received from its neighbor ach node keeps track of link state information received from other nodes ach node uses above information to determine next hop to each destination 14 7

8 xample ava applet: 15 istance Vector and ink tate in Wireless etworks ink state o ully network topology must be distributed throughout the network o an lead to short terms loop o OP elected router to solve distribution issues ot practical in wireless: mobility induces lots of topology changes, need to flood information throughout the network istance vector o onitor neighbors and outgoing links o Periodically broadcast shortest distance table to every neighbors ot practical in wireless: periodic update wastes bandwidth/power, links/connectivity time-varying, solutions to prevent loops are not directly applicable 16 8

9 Problems of conventional routing algorithms in wireless networks (RP, OP) are not designed for dynamic topologies inks are asymmetric any links available Periodic information (route,topology) update or dump broadcast o andwidth o Power 17 Routing in Wireless etworks: isclaimer! There is not a unique solution! o o one size fits all ots of different algorithms and proposals o ifferent design goals and objectives 18 9

10 Routing in T: ssumptions ully symmetric environment ll nodes/stations have identical capabilities and responsabilities 19 Route iscovery and aintenance (control plane): Proactive vs Reactive Proactive protocols o etermine routes independent of traffic pattern o Traditional link-state and distance-vector routing protocols are proactive Reactive protocols o aintain routes only if needed ybrid protocols 20 10

11 Reactive vs Proactive: Trade-Off atency of route discovery o Proactive protocols may have lower latency since routes are maintained at all times o Reactive protocols may have higher latency because a route from X to will be found only when X attempts to send to Overhead of route discovery/maintenance o Reactive protocols may have lower overhead since routes are determined only if needed o Proactive protocols can (but not necessarily) result in higher overhead due to continuous route updating Which approach achieves a better trade-off depends on the traffic and mobility patterns 21 Taxonomy of Routing Protocols for Wireless etworks (obile) ad hoc networks Response time, bandwidth ensor networks nergy Proactive protocols Reactive protocols estination-equenced istance-vector (V) Optimized ink- tate Routing (OR) d oc On-emand istance-vector (OV) ynamic ource Routing (R) eographybased routing (or hierarchical) luster-based routing 22 11

12 looding The wireless channel is a broadcast medium after all 23 looding for ata elivery ender broadcasts data packet P to all its neighbors ach node receiving P forwards P to its neighbors equence numbers used to avoid the possibility of forwarding the same packet more than once Packet P reaches destination provided that is reachable from sender ode does not forward the packet 24 12

13 looding for ata elivery Represents a node that has received packet P Represents that connected nodes are within each other s transmission range 25 looding for ata elivery roadcast transmission Represents a node that receives packet P for the first time Represents transmission of packet P 26 13

14 looding for ata elivery ode receives packet P from two neighbors: potential for collision 27 looding for ata elivery ode receives packet P from and, but does not forward it again, because node has already forwarded packet P once 28 14

15 looding for ata elivery odes and both broadcast packet P to node ince nodes and are hidden from each other, their transmissions may collide Packet P may not be delivered to node at all, despite the use of flooding 29 looding for ata elivery ode does not forward packet P, because node is the intended destination of packet P 30 15

16 looding for ata elivery looding completed odes unreachable from do not receive packet P (e.g., node ) odes for which all paths from go through the destination also do not receive packet P (example: node ) 31 looding for ata elivery looding may deliver packets to too many nodes (in the worst case, all nodes reachable from sender may receive the packet) 32 16

17 looding for ata elivery: dvantages implicity ay be more efficient than other protocols when rate of information transmission is low enough that the overhead of explicit route discovery/maintenance incurred by other protocols is relatively higher o this scenario may occur, for instance, when nodes transmit small data packets relatively infrequently, and many topology changes occur between consecutive packet transmissions Potentially higher reliability of data delivery o ecause packets may be delivered to the destination on multiple paths o (we will talk about it later) 33 looding for ata elivery: isadvantages Potentially, very high overhead o ata packets may be delivered to too many nodes who do not need to receive them Potentially lower reliability of data delivery o looding uses broadcasting -- hard to implement reliable broadcast delivery without significantly increasing overhead o roadcasting in is unreliable o n our example, nodes and may transmit to node simultaneously, resulting in loss of the packet o in this case, destination would not receive the packet at all 34 17

18 looding of ontrol Packets any protocols perform (potentially limited) flooding of control packets, instead of data packets The control packets are used to discover routes iscovered routes are subsequently used to send data packet(s) Overhead of control packet flooding is amortized over data packets transmitted between consecutive control packet floods 35 R: ynamic ource Routing R 4728: avid. ohnson. Routing in d oc etworks of obile osts. Proceedings of the Workshop on obile omputing ystems and pplications, pp , omputer ociety, anta ruz,, ecember 1994 Reactive protocol o ource routing for data delivery o Route discovery / route maintenance 36 18

19 ynamic ource Routing When node wants to send a packet to node, but does not know a route to, node initiates a route discovery ource node floods Route Request (RRQ) ach node appends own identifier when forwarding RRQ 37 Route iscovery in R Represents a node that has received RRQ for from 38 19

20 Route iscovery in R roadcast transmission [] Represents transmission of RRQ [X,] Represents list of identifiers appended to RRQ 39 Route iscovery in R [,] [,] ode receives packet RRQ from two neighbors: potential for collision 40 20

21 Route iscovery in R [,,] [,,] ode receives RRQ from and, but does not forward it again, because node has already forwarded RRQ once 41 Route iscovery in R [,,,] [,,,] odes and both broadcast RRQ to node ince nodes and are hidden from each other, their transmissions may collide 42 21

22 Route iscovery in R [,,,,] ode does not forward RRQ, because node is the intended target of the route discovery 43 Route iscovery in R estination on receiving the first RRQ, sends a Route Reply (RRP) RRP is sent on a route obtained by reversing the route appended to received RRQ RRP includes the route from to on which RRQ was received by node 44 22

23 Route Reply in R RRP [,,,,] Represents RRP control message 45 Route Reply in R Route Reply can be sent by reversing the route in Route Request (RRQ) only if links are guaranteed to be bidirectional o To ensure this, RRQ should be forwarded only if it received on a link that is known to be bi-directional f unidirectional (asymmetric) links are allowed, then RRP may need a route discovery for from node o Unless node already knows a route to node o f a route discovery is initiated by for a route to, then the Route Reply is piggybacked on the Route Request from. f is used to send data, then links have to be bi-directional (since ck is used) 46 23

24 ynamic ource Routing (R) ode on receiving RRP, caches the route included in the RRP When node sends a data packet to, the entire route is included in the packet header o hence the name source routing ntermediate nodes use the source route included in a packet to determine to whom a packet should be forwarded 47 ata elivery in R T [,,,,] Packet header size grows with route length 48 24

25 When to Perform a Route iscovery When node wants to send data to node, but does not know a valid route node 49 R Optimization: Route aching ach node caches a new route it learns by any means When node finds route [,,,,] to node, node also learns route [,,] to node When node receives Route Request [,,] destined for node, node learns route [,,,] to node When node forwards Route Reply RRP [,,,,], node learns route [,,] to node When node forwards ata [,,,,] it learns route [,,,] to node node may also learn a route when it overhears ata packets 50 25

26 Use of Route aching When node learns that a route to node is broken, it uses another route from its local cache, if such a route to exists in its cache. Otherwise, node initiates route discovery by sending a route request ode X on receiving a Route Request for some node can send a Route Reply if node X knows a route to node Use of route cache o an speed up route discovery o an reduce propagation of route requests 51 Use of Route aching [,,,,] [,,,] [,,],[,,] [,] [,,] [,,,] [P,Q,R] Represents cached route at a node (R maintains the cached routes in a tree format) 52 26

27 Use of Route aching: an peed up Route iscovery [,,,,] [,,,] [,,],[,,] [,] When node sends a route request for node, node sends back a route reply [,,,] to node using a locally cached route [,,] [,,,] RRQ [,,,] RRP 53 Use of Route aching: an Reduce Propagation of Route Requests [,,,,] [,,,] [,,],[,,] [,] [,,,] [,,] [,,,] RRP RRQ ssume that there is no link between and. Route Reply (RRP) from node limits flooding of RRQ. n general, the reduction may be less dramatic

28 Route aintenance f node is moving: can reinitiate route discovery f a link breaks: Route rror packets 55 Route rror (RRR) RRR [-] sends a route error to along route --- when its attempt to forward the data packet (with route ) on - fails odes hearing RRR update their route cache to remove link

29 Route aching: eware! tale caches can adversely affect performance With passage of time and host mobility, cached routes may become invalid sender host may try several stale routes (obtained from local cache, or replied from cache by other nodes), before finding a good route 57 ynamic ource Routing: dvantages Routes maintained only between nodes who need to communicate o Reduces overhead of route maintenance Route caching can further reduce route discovery overhead single route discovery may yield many routes to the destination, due to intermediate nodes replying from local caches 58 29

30 ynamic ource Routing: isadvantages Packet header size grows with route length due to source routing lood of route requests may potentially reach all nodes in the network are must be taken to avoid collisions between route requests propagated by neighboring nodes o nsertion of random delays before forwarding RRQ ncreased contention if too many route replies come back due to nodes replying using their local cache o Route Reply torm problem o Reply storm may be eased by preventing a node from sending RRP if it hears another RRP with a shorter route 59 ynamic ource Routing: isadvantages elay for route establishement Route maintenance is not a local process n intermediate node may send Route Reply using a stale cached route, thus polluting other caches This problem can be eased if some mechanism to purge (potentially) invalid cached routes is incorporated. or some proposals for cache invalidation, see [u00obicom] o tatic timeouts o daptive timeouts based on link stability 60 30

31 looding/roadcast torm ee.-. i,.-. Tseng,.-. hen, and.-p. heu, "The broadcast storm problem in a mobile ad hoc network," in obiom roadcast torm Problem When node broadcasts a route query, nodes and both receive it and both forward to their neighbors and transmit at about the same time since they are reacting to receipt of the same message from This results in a high probability of collisions 62 31

32 roadcast torm Problem Redundancy: given node may receive the same route request from too many nodes, when one copy would have sufficed ode may receive from nodes and both 63 olutions for roadcast torm Probabilistic scheme: On receiving a route request for the first time, a node will re-broadcast (forward) the request with probability p lso, re-broadcasts by different nodes should be staggered by using a collision avoidance technique (wait a random delay when channel is idle) o This would reduce the probability that nodes and would forward a packet simultaneously in the previous example 64 32

33 olutions for roadcast torms ounter-ased cheme: f node hears more than k neighbors broadcasting a given route request, before it can itself forward it, then node will not forward the request ntuition: k neighbors together have probably already forwarded the request to all of s neighbors 65 olutions for roadcast torms istance-ased cheme: f node hears RRQ broadcasted by some node within physical distance d, then will not re-broadcast the request ntuition: and are too close, so transmission areas covered by and are not very different o if re-broadcasts the request, not many nodes who have not already heard the request from will hear the request <d 66 33

34 olutions for roadcast torms et the physical layer characteristics handle it o igh packet loss rate 67 ummary: roadcast torm Problem looding is used in many protocols, such as ynamic ource Routing (R) Problems associated with flooding o ollisions o Redundancy ollisions may be reduced by jittering (waiting for a random interval before propagating the flood) Redundancy may be reduced by selectively rebroadcasting packets from only a subset of the nodes 68 34

35 d oc On-emand istance Vector Routing (OV) R 3561: harles. Perkins and lizabeth. Royer. "d hoc On-emand istance Vector Routing." Proceedings of the 2nd Workshop on obile omputing ystems and pplications, 1999 Reactive protocol o istance vector o estination sequence number 69 Why OV? R includes source routes in packet headers Resulting large headers can sometimes degrade performance o particularly when data contents of a packet are small OV attempts to improve on R by maintaining routing tables at the nodes, so that data packets do not have to contain routes OV retains the desirable feature of R that routes are maintained only between nodes which need to communicate 70 35

36 OV Overview Route Requests (RRQ) are forwarded in a manner similar to R When a node re-broadcasts a Route Request, it sets up a reverse path pointing towards the source o OV assumes symmetric (bi-directional) links When the intended destination receives a Route Request, it replies by sending a Route Reply Route Reply travels along the reverse path set-up when Route Request is forwarded 71 OV overview Path discovery Reverse path setup orward path setup 72 36

37 Route Requests in OV Represents a node that has received RRQ for from 73 Route Requests in OV roadcast transmission Represents transmission of RRQ 74 37

38 Route Requests in OV Represents links on Reverse Path 75 Reverse Path etup in OV ode receives RRQ from and, but does not forward it again, because node has already forwarded RRQ once 76 38

39 39 Reverse Path etup in OV 77 Reverse Path etup in OV ode does not forward RRQ, because node is the intended target of the RRQ 78

40 Route Reply in OV Represents links on path taken by RRP 79 Route Reply in OV n intermediate node (not the destination) may also send a Route Reply (RRP) provided that it knows a more recent path than the one previously known to sender To determine whether the path known to an intermediate node is more recent, destination sequence numbers are used The likelihood that an intermediate node will send a Route Reply when using OV not as high as R o new Route Request by node for a destination is assigned a higher destination sequence number. n intermediate node which knows a route, but with a smaller sequence number, cannot send Route Reply 80 40

41 orward Path etup in OV orward links are setup when RRP travels along the reverse path Represents a link on the forward path 81 ata elivery in OV T Routing table entries used to forward data packet. Route is not included in packet header

42 Timeouts routing table entry maintaining a reverse path is purged after a timeout interval o Timeout should be long enough to allow RRP to come back o Used to remove unused reverse path routing table entry maintaining a forward path is purged if not used for a active_route_timeout interval o f no data is being sent using a particular routing table entry, that entry will be deleted from the routing table (even if the route may actually still be valid) 83 ink ailure Reporting neighbor of node X is considered active for a routing table entry if the neighbor sent a packet within active_route_timeout interval which was forwarded using that entry When the next hop link in a routing table entry breaks, all active neighbors are informed ink failures are propagated by means of Route rror messages, which also update destination sequence numbers 84 42

43 Route rror When node X is unable to forward packet P (from node to node ) on link (X,), it generates a RRR message ode X increments the destination sequence number for cached at node X The incremented sequence number is included in the RRR When node receives the RRR, it initiates a new route discovery for using destination sequence number at least as large as 85 estination equence umber ontinuing from the previous slide When node receives the route request with destination sequence number, node will set its sequence number to, unless it is already larger than 86 43

44 ink ailure etection ello messages: eighboring nodes periodically exchange hello message bsence of hello message is used as an indication of link failure lternatively, failure to receive several -level acknowledgement may be used as an indication of link failure 87 Why equence umbers in OV To avoid using old/broken routes o To determine which route is newer To prevent formation of loops o ssume that does not know about failure of link - because RRR sent by is lost o ow performs a route discovery for. ode receives the RRQ (say, via path --) o ode will reply since knows a route to via node o Results in a loop (for instance, ---- ) 88 44

45 Why equence umbers in OV o oop Optimization: xpanding Ring earch Route Requests are initially sent with small Time-to- ive (TT) field, to limit their propagation o R also includes a similar optimization f no Route Reply is received, then larger TT tried 90 45

46 ummary: OV Routes need not be included in packet headers odes maintain routing tables containing entries only for routes that are in active use t most one next-hop per destination maintained at each node o ulti-path extensions can be designed o R may maintain several routes for a single destination Unused routes expire even if topology does not change 91 ixing istance Vector in Wireless etworks: V V: destination sequence distance vector dapting distance vector algorithms for wireless network olves routing loop problem Perkins, harles. and hagwat, Pravin. ighly ynamic estination-equenced istance-vector Routing (V) for obile omputers, igcomm

47 V Overview ach node maintains a routing table which stores o ext hop towards each destination o cost metric for the path to each destination o destination sequence number that is created by the destination itself o equence numbers used to avoid formation of loops ach node periodically forwards the routing table to its neighbors o ach node increments and appends its sequence number when sending its local routing table o This sequence number will be attached to route entries created for this node 93 V Overview Routing information distribution o ull dumps infrequently o ncremental updates (smaller) more frequently 94 47

48 V: equence umber ssume that node X receives routing information from about a route to node X et (X) and () denote the destination sequence number for node as stored at node X, and as sent by node with its routing table to node X, respectively 95 V: equence umber and Routing Update ode X takes the following steps: X o f (X) > (), then X ignores the routing information received from o f (X) = (), and cost of going through is smaller than the route known to X, then X sets as the next hop to o f (X) < (), then X sets as the next hop to, and (X) is updated to equal () 96 48

49 ummary ll protocols discussed so far perform some form of flooding o R, OV o V Use hop count metric re not connected to the nternet 97 49