A Highly Adaptive Distributed Routing

Algorithm for Mobile Wireless Networks

Research Paper

By

V. D. Park and M. S. Corson

Introduction A distributed routing protocol for mobile,

multihop, wireless networks is presented Protocol is a type of “link reversal” algorithms The protocol is highly adaptive, efficient and

scalable Best suited for dense networks

Protocol uses a synchronized physical or logical clock

Temporally Ordered Routing Algorithm (TORA)

Introduction A mobile adhoc network is a collection of

mobile routers Mobility requires fast protocol adaptation for

routing purposes Existing shortest path algorithms and

adaptive shortest path algorithms are not well suited in a dense and highly dynamic environment These algorithm maintain only one path

Routing in Mobile Networks Access Point Configuration

The access point does the routing Everyone node is at one hop distance No routing

Adhoc Network Every node is a router Nodes can be at 1 – N hops

MANET The complete network becomes mobile

Routing in Mobile Networks

N1

AP

N3

N2

N1 N3

N2

N4 N5

Routing in Mobile Networks

N1

AP

N3

N2

N1 N3

N2

N4 N5

Existing Protocols Some Existing algorithms include

Gafini Bertsekas (GB) algorithm Lightweight Mobile Routing (LMR) protocol Destination Sequenced Distance Vector (DSDV) Wireless Routing Protocol (WRP) Dynamic Source Routing (DSR) protocol

All these algorithms have stability problem in dense networks or are limited by discovery of only one path

Efficient Routing in Dense Mobile Networks

Efficient routing in dense networks require Distributed execution Loop free routing Provision of multiple routes Quick route establishment Less communication overhead

Routing optimality is of less significance sine mobility changes the shortest path

TORA Protocol Assumptions

Each node N has a unique identifier (ID) Each link L allows two-way communication A node failure occurs if all links incident to that node are severed Each node is always aware of its neighbors All transmission are broadcasted and transmissions are received in order All nodes have synchronized clocks

Notations Network is modeled as directed acyclic graph

N is a finite set of nodes L is a set of initially undirected links

Each link L maybe assigned one of the following states Undirected Directed from node i to j - upstream Directed from node j to i - downstream

),( LNG

Foundation and Basic Structure Creating routes

Establishment of a sequence of directed links leading from node to destination

Only initiated when a node with no directed links require a route through query/reply messages

Assigning direction to links in n undirected network or portion Maintaining routes

Reacting to topological changes in a network in a manner such that routes to the destination are established within a finite time

The directed portions returned to a destination oriented DAG Erasing routes

Upon detection of a network partition all links must be undirected to erase invalid routes

Foundation and Basic Structure Protocol uses three distinct control packets

Query (QRY) Creating routes

Update (UPD) Creating and maintaining routes

Clear (CLR) Erasing routes

Algorithm Description At any given time an ordered quintuple is associated

with each node

First value is a time tag set to the time of link failure Second value is a unique originator ID (unique ID of the node

which defined the new reference level) Reference levels can be ordered lexicographically

Third value is a single bit used to divide each of the unique reference levels into two unique sub-level

Fourth value is an integer used to order nodes with respect to a common reference level

last value is the unique ID of the node itself

),,,,( iroidH iiiii

Ni

Algorithm Description Conceptually the quintuple represents the

height of the node with respect to A reference level represented by the first three

values in the quintuple A new reference level is defined each time a node

loses its last downstream link due to a link failure A delta value instrumental in propagation of a

reference level

Route Creation Routes are created using QRY and UPD

packets QRY consists of a destination ID (did) UPD consists of a did and the height of the node

i which is broadcasting the packet H Each node other than destination maintains a

Route-Required (RR) flag which is initially un-set and the time at which last UPD packet was broadcast and the time at which each link became active

Route Creation When QRY is Received If the receiving node has no downstream links and RR is

un-set it re-broadcasts the QRY and sets the RR If the receiving node has no downstream links and its RR

is set it discard the QRY packet If receiving node has atleast one downstream link and its

height is NULL it sets its height and broadcasts a UPD (delta value is incremented)

If the receiving node has atleast one downstream link and its height is non-NULL

It first compares the time last UPD was broadcasted to the time the link over which the QRY packet was received became active

If UPD has been broadcast since the link became active it discards QRY otherwise, broadcasts UPD packet. If a node has the RR flag set when new link is established it broadcasts a QRY packet

Route Creation

Maintaining Routes Performed only for nodes having a height

other than NULL Any neighbor having a NULL height is not used

Maintaining RoutesNode i loses its last

downstream link

Was the linklost due to a

failure?Case 1

Generate newreference level

Case 2Propagate the

highest neighbor’sreference level

Do all the neighbors havethe same reference level?

Is the reflection bit inthat reference set to 1?

Did this node originallydefine that reference level?

Case 3Reflect back a

higher sub level

Case 4Partition detected,

erase invalid routes

Case 5Generate newreference level

YES NO

NO

YES

NOYES

YESNO

Maintaining Routes

Maintaining Routes

Erasing Routes A propagation of CLR packet erases routes

CLR is broadcasted upon the detection of a partition in the network

Erasing Routes

Performance No comparative simulation results are

available (at the time of writing of the paper)

Complexity comparison with other protocols show that TORA has linear complexity

Complexity Comparison

Conclusion A highly adaptive distributed routing

algorithm is proposed Suited for operation in mobile wireless networks

Maintains loop free multipath routing to destinations for which routing is required

Transmits three type of control messages for creation, updation and deletion of routes