Ad-Hoc Networks: Routing Algorithms

Advanced Algorithms & Data Structures Lecture Theme 05 Robin Pomplun Summer Semester 2006

Overview

Mobile Ad Hoc Networks Position-unaware Routing Algorithms

Proactive Protocols Destination-Sequenced Distance Vector Protocol Reactive Protocols Dynamic Source Routing Protocol Ad-Hoc On-Demand Routing Protocol Hybrid Protocols Zone Routing Protocol Research

Mobile Ad Hoc Networks

Definition: A mobile ad hoc network is an autonomous system of mobile routers connected by wireless links the union of which form an arbitrary graph. The routers are free to move randomly and organize themselves arbitrarily; thus, the network's wireless topology may change rapidly and unpredictably.
Source: Internet Engineering Task Force (IETF)

An Example

Networking without an Infrastructure

If all the wireless nodes are within the transmission range of each other, routing is easy. Every node can listen to all transmissions. However, this is not true in most situations, due to short transmission range. Hence, most ad hoc networks are multi-hop. A message from a source node must then go through intermediate nodes to reach its destination.

An Example

Position-unaware Routing
We will assume that each node is capable of running fairly complicated algorithms locally. Each node has an address and each node keeps track of the neighbours addresses. However, when a message has to be sent from a source to a destination, the source node is not aware of the current position of the destination.

Position-unaware Routing
Position-unaware Routing Protocols can be classified based on the way a protocol tries to find a route to a destination: Proactive Routing Protocol Reactive Routing Protocol

Proactive Routing
Entire network topology is known to all nodes and maintained in a routing table Since each node knows the complete topology, a node can immediately find the best route to a destination. Routing messages are exchanged among the nodes periodically to update their routing tables Routing Table Advertising

Protocols: Destination-Sequenced Distance Vector (DSDV) Fisheye State Routing (FSR) ...
9

Destination-Sequenced Distance Vector Protocol

Packets are transmitted between the nodes using route tables stored at each node. Each route table lists all available destinations and the number of hops to each destination. For each destination, a node knows which of its neighbours leads to the shortest path to the destination.

10

Routing Table Entries

The destinations address Next Hop to Destination The number of hops to the destination The sequence number of the information received from that destination. This is the original sequence number assigned by the destination.

n m

11

How the local Routing Table is Used

k l i m 8 23 x 13

Consider a node i. Suppose, i needs to send a message to node x. i can look up the best route to x from its routing table and forwards the message to the neighbour along the best route. The neighbour in turn checks the best route from its own table and forwards the 12 message to its appropriate neighbour. The routing progresses this way.

Routing Table Advertising

The DSDV protocol requires each mobile node to advertise its own route table to all of its current neighbours. Each mobile node agrees to forward route advertising messages from other mobile nodes. This forwarding is necessary to send the advertisement messages all over the network. In other words, route advertisement messages help mobile nodes to get an overall picture of the topology of the network.

13

Responding to Topology Changes

It is necessary to avoid excessive control traffic (route update information). Otherwise, the bandwidth will be taken up by control traffic. The solution is to broadcast two types of updates: Full Dump / Incremental Dump A full dump carries complete routing tables. A node broadcasts a full dump infrequently. An incremental dump carries minor changes in the routing table. This information contains changes since the last full dump. When the size of an incremental dump becomes too large, a full dump is preferred.
14

Responding to Topology Changes

When a node i receives incremental dump or full dump from another node j, the following actions are taken : The sequence number of the current dump from j is compared with previous dumps from j If the sequence number is new, the route table at i is updated with this new information. (Reason for Sequence number: Loops) Node i now broadcasts its new route table as an incremental or a full dump.

15

An Example of Route Update

At the start, each node gets route updates only from its neighbours. For n4, the distances to the other nodes are : n5=1, n3=1, n2= n1 =

n1 n2 n3 n4 n5

All nodes broadcast with a sequence number 1

16

An Example of Route Update

After this, nodes forward messages that they have received earlier. The message that n2 sent to n3 is now forwarded by n3 For n4, the distances are now : n5=1, n3=1, n2=2, n1=

n1 n2 n3 n4 n5

All messages have sequence number 1

17

An Example of Route Update

Finally, after second round of forwarding, n4 gets the following distances : n5=1, n3=1, n2=2, n1=3

n1 n2 n3 n4 n5

18

An Example of Route Update

Suppose n5 has moved to its new location. Also, n5 receives a new message from n1 with a sequence number 2 This message is forwarded by n5 to n4 Two distances to n1 in n4 Distance 3 with sequence number 1, and Distance 2 with sequence number 2 Since the latter message has a more recent sequence number, n4 will update the distance to n1 as 2

n1

n5

n2 n3 n4 n5

19

How good is DSDV?

DSDV is an efficient protocol for route discovery. Whenever a route to a new destination is required, it already exists at the source. Hence, latency for route discovery is very low. However, DSDV needs to send a lot of control messages. These messages are important for maintaining the network topology at each node. This may generate high volume of traffic for high-density and highly mobile networks.

20

Performance Evaluation
Performance Evaluation with real networks is to expensive Freely available Simulation Tool (e.g. Network Simulator ns-2) Simulation is done on packet level (trace file) Generate random mobility Random Waypoint Model Generate Traffic (TCP, UDP)

21

Trace File (ns-2)

s -t 0.500000000 -Hs 0 -Hd -2 -Ni 0 -Nx 16.77 -Ny 16.77 -Nz 0.00 -Ne -1.000000 -Nl AGT -Nw --- -Ma 0 -Md 0 -Ms 0 -Mt 0 -Is 0.0 -Id 1.0 -It tcp -Il 40 -If 2 -Ii 0 -Iv 32 -Pn tcp -Ps 0 -Pa 0 -Pf 0 -Po 0 r -t 0.500000000 -Hs 0 -Hd -2 -Ni 0 -Nx 16.77 -Ny 16.77 -Nz 0.00 -Ne -1.000000 -Nl RTR -Nw --- -Ma 0 -Md 0 -Ms 0 -Mt 0 -Is 0.0 -Id 1.0 -It tcp -Il 40 -If 2 -Ii 0 -Iv 32 -Pn tcp -Ps 0 -Pa 0 -Pf 0 -Po 0 s -t 0.506923068 -Hs 0 -Hd -1 -Ni 0 -Nx 16.79 -Ny 16.79 -Nz 0.00 -Ne -1.000000 -Nl RTR -Nw --- -Ma 0 -Md 0 -Ms 0 -Mt 0 -Is 0.255 -Id 1.255 -It DSR -Il 32 -If 0 -Ii 1 -Iv 32 -P dsr -Ph 1 -Pq 1 -Ps 1 -Pp 0 -Pn 1 -Pl 0 -Pe 0->0 -Pw 0 -Pm 0 -Pc 0 -Pb 0->0 s -t 0.508398068 -Hs 0 -Hd -1 -Ni 0 -Nx 16.80 -Ny 16.80 -Nz 0.00 -Ne -1.000000 -Nl MAC -Nw --- -Ma 0 -Md 0 -Ms ffff0008 -Mt 0 -Is 0.255 -Id 1.255 -It DSR -Il 32 -If 0 -Ii 1 -Iv 32 -P dsr -Ph 1 -Pq 1 -Ps 1 -Pp 0 -Pn 1 -Pl 0 -Pe 0->0 -Pw 0 -Pm 0 Pc 0 -Pb 0->0 r -t 0.508526679 -Hs 1 -Hd -1 -Ni 1 -Nx 146.78 -Ny 146.06 -Nz 0.00 -Ne -1.000000 Nl MAC -Nw --- -Ma 0 -Md 0 -Ms ffff0008 -Mt 0 -Is 0.255 -Id 1.255 -It DSR -Il 32 -If 0 -Ii 1 -Iv 32 -P dsr -Ph 1 -Pq 1 -Ps 1 -Pp 0 -Pn 1 -Pl 0 -Pe 0->0 -Pw 0 -Pm 0 -Pc 0 -Pb 0->0
22

Performance Evaluation
Performance Evaluation with real networks is to expensive Freely available Simulation Tool (e.g. Network Simulator ns-2) Simulation is done on packet level (trace file) Generate random mobility Random Waypoint Model Generate Traffic (TCP, UDP)

23

Evaluation
Packet Delivery Ratio:
The ratio of the data packets delivered to the destinations to the number of sent packets

Average End-End-Delay (Latency):

This includes all possible delays caused by buffering during route discovery, queuing at the interface queue, retransmission delays, propagation and transfer times.

Routing Overhead Number of all packets which are needed to do Route / Topology Maintenance Throughput (in Bits/sec or Packets/sec) Number of Bits (Packets) per second which are received in the network Scenario: Number of Nodes: 50 Transmission Range: 250m Plane: 500x500m Simulation Time: 200 (start: 100)
24

Evaluation

25

Reactive Routing
In a reactive protocol, a route is discovered only on-demand, when it is necessary. These protocols generate much less control traffic at the cost of latency, i.e., it usually takes more time to find a route compared to a proactive protocol.

Protocols: Dynamic Source Routing (DSR) Ad Hoc On-Demand Distance-Vector (AODV) ...

26

Dynamic Source Routing

Each node maintains a route cache to remember routes that it has learnt about. A node may store multiple routes to a destination in its route cache. A node can react to changes in network topology much more rapidly by taking advantage of cached routes. For example, if one route to a destination is broken, the source node can choose another route to the destination from its route cache.

27

Route Discovery
The DSR protocol has two important mechanisms through which the protocol operates. Route Discovery: A node A wishing to send a packet to node E obtains a route to E Route Request Route Reply Route Maintenance: When A is using a discovered route to E, A may detect that the route is broken. In such cases, A may use an alternate route to E (if it is known), or start another route discovery phase to E.

28

Route Discovery
Node A is trying to discover a route to node E. A broadcasts a route request message to its neighbours. This message is received by all nodes within the transmission range of A. Each route request message contains the source and target of the route discovery. Also, each route request is stamped with an unique ID assigned by the source.

D A B A C AB ABC E ABCD

29

Route Discovery
Every node that receives a route request message, does one of the following: Check the unique ID of route request; already received: discard RReq A node like C first searches its route cache to see whether it has a stored route to E. If it has such a route, C sends that route to A. (Route Reply) If there is no such route in its route cache, C broadcasts the route request message to its neighbours. C attaches its own ID to the route request message

A B A

AB C ABC

ABCD E

30

Example

31

Route Maintenance
The DSR protocol has two important mechanisms through which the protocol operates. Route Discovery: A node A wishing to send a packet to node E obtains a route to E Route Request Route Reply Route Maintenance: When A is using a discovered route to E, A may detect that the route is broken. In such cases, A may use an alternate route to E (if it is known), or start another route discovery phase to E.

32

Route Error
B A C D E

A node like C tries to forward the message and waits for acknowledgment. C will retransmit the message a fixed number of times if no acknowledgment arrives. After that, C will initiate a route error message. In this example, C will initiate a route error message back to A indicating that the link to D is currently broken. A will remove this route from its route cache and try another route to E, if it has one. Or, A may start a new route discovery.
33

Caching Overheard Routing Information

DSR extensively takes advantage of existing knowledge of the network topology. Each node gathers information about the network topology by overhearing other nodes transmissions.

C A B D

P
34

Route Request Hop Limit

Sometime it is not good to propagate a route request message throughout the network. In case D is in the neighbourhood of S, the route request message from S should not propagate too far away. If D is near S, propagating the route request message too far will result in too many unnecessary route reply messages in future.

D
35

Restricted Propagation of Route Request

A better strategy is to propagate route request messages with increasing hop count. Initially, send the route request to a distance of 2 hops. If no route reply is received after sometime, send the route request to a distance of 4 hops and so on. This reduces network congestion by reducing the number of route reply messages.

36

Non-uniform Packet Size in DSR

When a source node A sends a packet to a destination node E, A should send the entire route to E along with the packet. This is necessary for the intermediate nodes to forward the packet. Usually all media support packets of uniform size. If a packet is large, it has to be split into smaller packets. This may cause problems in the wireless medium as packets that are split into smaller parts may not arrive in correct order. Intermediate nodes may not be able to forward packets correctly.

37

Performance

38

Summary DSDV & DSR

DSDV Advantages: Low Latency (all information saved in the routing tables) Disadvantages: High Overhead (local movements have global effect) Independent of traffic (no traffic, same overhead) DSR Advantages: On-Demand (only routing overhead if traffic) Disadvantages: High Latency (no route in cache, route discovery) Non-Uniform Packet Size (in case of long routes very bad)
39

Summary DSDV & DSR

40

Ad-Hoc On-Demand Distance-Vector (AODV)

Table-Driven Routing Tables (for saving information about topology) On-Demand Route Discovery Expanding Ring Search (Route Request Type) Forward Path Setup (Saving Route Information in RTables) Route Maintenance (Methods to repair broken links)

41

Routing Tables

Entries: Destination IP Next Hop IP Destination Sequence Number Life Time List of Precursors Hop Count Number Sequence Number: The Seq. Number is monotonically increased each time the node learns of a change in the topology.

42

Example (Routing Table)

A B C D

F E
Routing Table of Node F:
Dest. IP B C D Next Hop IP B G G Dest. SeqNo. Lifetime 2 4 4 10 6 8 Precursors E E A, E Hop Count 1 2 2
43

Route Discovery

Like DSR, we use two types of messages, route request (RREQ) and route reply (RREP): Route Request Messages (RREQ) Route Request Processing (RREQ) Reverse Route Entry Expanding Ring Search (RREQ) Responding to Route Request Messages (RREP) Forward Path Setup

44

Route Request Message

When node S wants to send a message to node D, S searches its routing table for a valid route to D. If there is no valid route, S initiates a RREQ message with the following components : The IP addresses of S and D The current sequence number of S and the last known sequence number of D A broadcast ID from S. This broadcast ID is incremented each time S initiates a RREQ message. Hop count The <broadcast ID, IP address> pair of the source S forms a unique identifier for the RREQ.

45

Example (Routing Request Message)

A B D E

F S
Routing Table of Node S:
Dest. IP D Next Hop IP F Dest. SeqNo. Lifetime 4 0 Precursors Hop Count 3

Routing Request Message:

Source IP S Dest. IP D Source Seq. No. 7 Dest. Seq. No. 4 Broadcast ID 15 Hop Count 0
46

Route Request Processing

Suppose a node P receives the RREQ from S. P first checks whether it has received this RREQ before. Each node stores the <broadcast ID, IPaddress> pairs for all the recent RREQs it has received. (for a specific amount of time)

S
47

Processing a RREQ Message

If P has seen this RREQ from S already, P discards the RREQ. Otherwise, P processes the RREQ : P sets up a reverse route entry in its routing table for the source S. This entry contains the IP address and current sequence number of S, number of hops to S and the address of the neighbour from whom P got the RREQ.
48

Reverse Route Entry

A D E

F S
Routing Table of Node G:
Dest. IP S Next Hop IP F Dest. SeqNo. Lifetime 7 10 Precursors Hop Count 2

Routing Request Message:

Source IP S Dest. IP D Source Seq. Dest. Seq. No. No. 7 4 Broadcast ID 15 Hop count 2
49

Responding to a RREQ Message

S

P can respond to the RREQ from S if P has an unexpired entry for D in its routing table. Moreover, the sequence number from D that P has, must not be less than the sequence number of D that was in the RREQ from S. This ensures that there is no loop in the route. If P satisfies both of these requirements, it sends a RREP message back to S. If P cannot reply to the RREQ from S, P increments the hop-count of the RREQ and broadcasts it to its neighbours.
50

Responding to a RREQ Message

A D E

F G S
Routing Table of Node G:
Dest. IP D Next Hop IP D Dest. SeqNo. Lifetime 6 10 Precursors X Hop Count 1

Routing Request Message:

Source IP S Dest. IP D Source Seq. Dest. Seq. No. No. 7 4 Broadcast ID 15 Hop Count 2
51

Forward Path Setup (Sending a RREP)

M S N RREP
A RREP message has several fields : The IP address of both source and destination If the destination is sending the RREP, it sends its current sequence number, a lifetime for the route and sets the hop-count to 0 If an intermediate node is responding, it sends the last known sequence number from the destination, sets the hop-count equal to distance from the destination and a lifetime for the route.

52

Responding to a RREP Message

A D E

F G S
Routing Table of Node G:
Dest. IP D Next Hop IP D Dest. SeqNo. Lifetime 6 4 Precursors X Hop Count 1

Routing Reply Message:

Source IP S Dest. IP D Source Seq. Dest. Seq. No. No. 7 6 Lifetime 4 Hop Count 1
53

Forward Path Setup

M S N RREP
A node (here Node M) sends a RREP back to a neighbour from whom it received the RREQ. When an intermediate node (here Node N) receives a RREP, it sets up a forward path to the destination in its route table. This contains the IP addresses of the neighbour and the destination, hopcount to the destination and a lifetime for the route.

54

Forward Path Setup

A D E

F S G
Dest.Seq. No Lifetime 7 6 10 4 Precursors G S Hop Count 1 2

Routing Table of Node F:

Dest. IP S D Next Hop IP S G

Routing Reply Message:

Source IP S Dest. IP D Source Seq. Dest. Seq. No. No. 7 6 Lifetime 4 Hop Count 1
55

Handling more than one RREP

M S P RREP D

An intermediate node P may receive more than one RREP for the same RREQ. P forwards the first RREP it receives and forwards a second RREP later only if : The later RREP contains a greater sequence number for the destination, (RREP for later RREQ) The hop-count to the destination is smaller in the later RREP Otherwise, it does not forward the later RREPs. This reduces the number of RREPs propagating towards the source.
56

Expanding Ring Search

For route discovery, a source node broadcasts a RREQ across the network. This may create a lot of messages in a large network. A source node uses an expanding ring search strategy. With a ring diameter K, a RREQ dies after its hop-count exceeds K. If a RREQ fails, the source node increases the value of K incrementally.

57

Example (Expanding Ring Search)

D

Routing Request Message:

Source IP S Dest. IP D Source Seq. No. 5 Dest. Seq. Broadcast No. Id 4 3 Hop Count 0 Time To Live 1
58

Route Maintenance
M S N D

Once a route has been established between two nodes S and D, it is maintained as long as S (source node) needs the route. If S moves during an active session, it can reinitiate route discovery to establish a new route to D. When D or an intermediate node moves, a route error (RERR) message is sent to S.

59

Route Error
RERR S

RERR 1 2

3 3 D

If S moves during an active session, it can reinitiate route discovery to establish a new route to D. When D or an intermediate node moves, a route error (RERR) message is sent to S.

Example: The link from node 3 to D is broken as 3 has moved away to a position 3. Node 2 sends a RERR message to 1 and 1 sends the message in turn to S. S initiates a route discovery if it still needs the route to D.
60

Updating Routing Tables

RERR S RERR 1 2 4

3 3 D 5

Suppose neighbours 4 and 5 route through 2 to reach D. Node 2 broadcasts RERR to all such neighbours. Each neighbour marks its route table entry to D as invalid by setting the distance to infinity. Each neighbour in turn propagates the RERR message.

61

Example (Routing Error)

A B C D

F E
Routing Table of Node F:
Dest. IP B C D Next Hop IP B G G Dest. SeqNo. Lifetime 2 4 4 10 6 8 Precursors E E A, E Hop Count 1 2 2
62

G G

Overview
AODV does not retransmit data packets that are lost and hence does not guarantee packet delivery. However, the packet delivery percentage is close to 100 with relatively small number of nodes. The packet delivery percentage drops with increased mobility. The overhead packets in AODV are due to RREQ, RREP and RERR messages. AODV needs much less number of overhead packets compared to DSDV. The number of overhead packets increases with increased mobility, since this gives rise to frequent link breaks and route discovery. The route discovery latency in AODV is low compared to DSR and DSDV.

63

AODV Evaluation

64

AODV vs. DSR & DSDV

65

Hybrid Routing Protocol

Proactive Protocols: Low Latency (Routing Tables map whole Network Topology) High Overload (Updating of Routing Tables) Reactive Protocols Low Overload (On-Demand Protocol) High Latency (Route Discovery needed) Idea: Mix reactive and proactive parts together Introduce Routing Zones for Zone Routing Protocol

66

Routing Zones
Routing Zone for Source Node S Routing Zone radius = 2 Peripheral Nodes: G, H, J, I, K

K B G S C H J E D A I

67

Basic Strategy of ZRP

Intrazone Routing: First, the packet is sent within the routing zone of the source node to reach the destination node or peripheral nodes. Interzone Routing: Then the packet is sent from the peripheral nodes towards the destination node.

D S

68

Intrazone - Routing
Each node collects information about all the nodes in its routing zone proactively. This strategy is similar to a proactive protocol like DSDV. Each node maintains a route table for its routing zone, so that it can find a route to any node in the routing zone from this table. Zone Notification Message (k-Hop Hello Message) to maintain Routing Table for Routing Zone

69

Interzone - Routing
The Interzone routing discovers routes to the destination reactively. Consider a source (S) and a destination (D). If D is within the routing zone of S, the routing is completed in the Intrazone routing phase. Otherwise, S sends the packet to the peripheral nodes of its zone through bordercasting. Bordercasting: S maintains complete routing table for its zone and routes the packet to the peripheral nodes by consulting this routing table.

D S

70

Interzone Route Discovery

S sends a route request (RREQ) message to the peripheral nodes of its zone through bordercasting. Each peripheral node P executes the same algorithm. First, P checks whether the destination D is within its routing zone and if so, sends the packet to D. Otherwise, P sends the packet to the peripheral nodes of its routing zone through bordercasting.

71

Example (Interzone Routing)

B D H

72

Evaluation
When the radius of the routing zone is 1, the behaviour of ZRP is like a pure reactive protocol, for example, like DSR. When the radius of the routing zone is infinity (or the diameter of the network), ZRP behaves like a pure proactive protocol, for example, like DSDV. The optimal zone radius depends on node mobility and route query rates / number of traffic connections.

73

Research

DSR with non-uniform packet size (intermediate nodes as routers) Mobility and Protocols (Traffic?) modeling of mobility Freeway / Manhattan / Pedestrian Scenario distribution of mobility every node has its own mobility environment which metric decides on mobility? (link duration?) what to do with information about mobility? change routing zone radius (ZRP) Network Scenarios large networks different mobility scenarios wireless plus connection point to internet through wired network

74

References
Charles E. Perkins (Hrsg.). Ad Hoc Networking, Addison-Wesley 2000. D. B. Johnson, D. A. Maltz, and Y. Hu. The dynamic source routing protocol for mobile ad hoc networks (DSR). Technical report, IETF MANET Working Group, 2004. Ch. E. Perkins, E. M. Belding-Royer, and S. R. Das. Ad hoc on-demand distance vector (AODV) routing. RFC Experimental 3561, Internet Engineering Task Force, July 2003. Ch. E. Perkins and P. Bhagwat. Highly dynamic destination-sequenced distancevector routing (DSDV) for mobile computers. In ACM Conference on Communications Architectures, Protocols and Applications, SIGCOMM '94, London, UK, pages 234-244. ACM, ACM, August 1994. Z. J. Haas, M. R. Pearlman, and P. Samar. The Zone Routing Protocol (ZRP) for ad hoc networks. Internet-draft, IETF MANET Working Group, July 2002. VINT. The network simulator - ns-2., http://www.isi.edu/nsnam/ns/ Lecture: Mobile Computing, Amitava Datta, http://electures.informatik.unifreiburg.de:8484/catalog/course.do?courseId=mcws200304
75