what is the way to determine if a route is a neighbor?

When that occurs any route using that computer as a next hop should be deleted including the route indicating that computer as the actual formerly neighboring destina tion Increasing the number of update periods that may transpire before entries are determined would result inmore stale routing entries but would also allow for more transmission errors Transmission errors are likely to occur when a CSMA type broadcast medium is used as may well be the case for many wireless implementa tions

What is the reason why the loop free property holds in the second scenario?

The loop free property holds in the second scenario due to the theorem proved in which states that in presence of static or decreasing link weights distance vector algorithms always maintain loop free pathsComparison with other Methods

What is the primary cause of loop formation in BF style algorithms?

The primary cause of loop formation in BF style algorithms is that nodes make uncoordinated modi ca tions to their routing tables based on some information which could be incorrect

how many different structures are used to represent the different types of routes?

In order to combat problems arising with large pop ulations of mobile hosts which can cause route updates to be received in an order delaying the best metrics until after poorer metric routes are received the authors have sepa rated the route tables into two distinct structures

What is the way to get rid of the uctuation problem?

To achieve this when there is a history of uctuations at Mobile HostMH the link breakage should be detected fast enough so that an intermediate host in Collection II nds out the problem and begins a triggered incremen tal update showing an metric for the path along the way to Mobile HostMH Routes with an metric are required by this protocol to be advertised immediately without delayIn order to bias the damping mechanism in favor of recent events the most recent measurement of the set tling time of a particular route must be counted with a higher weighting factor than are less recent measure ments

What is the main concern of ad hoc routing?

The primary concern with using a Distributed Bell man Ford algorithm in ad hoc environment is its suscep tibility towards forming routing loops and counting to in nity problem RIP which is very similar to DBF algorithm also su ers from the same problem Unlike DBF RIP only keeps track of the best route to each destination which results in some space saving at no extra performance hit RIP also employs techniques known as split horizon and poisoned reverse to avoid a ping pong style of looping but these techniques are not powerful enough to avoid loops involvingmore than two hops

(Open Access) Highly dynamic Destination-Sequenced Distance-Vector routing (DSDV) for mobile computers (1994) | Charles E. Perkins

Q: What contributions have the authors mentioned in the paper "Highly dynamic destination sequenced distance vector routing dsdv for mobile computers" ?

In this paper the authors present an innovative design for the operation of such ad hoc networks The authors have investigated modi cations to the basic Bellman Ford routing mechanisms as speci ed by RIP to make it suitable for a dynamic and self starting network mechanism as is required by users wishing to utilize ad hoc networks Their modi cations address some of the previous objections to the use of Bellman Ford related to the poor looping properties of such algorithms in the face of broken links and the resulting time dependent nature of the interconnection topology describing the links between the Mobile Hosts Finally the authors describe the ways in which the basic network layer routing can be modi ed to provide MAC layer support for ad hoc networks

Highly Dynamic Destination-Sequenced Distance-Vector Routing

(DSDV) for Mobile Computers

Charles E. Perkins

IBM, T.J. Watson ResearchCenter

Hawthorne, NY 10562

Pravin Bhagwat

Computer Science Department

University of Maryland

College Park, MD 20742

Abstract

ad-hoc

network is the cooperative engagementof

a collection of Mobile Hosts without the required inter-

vention of anycentralized Access Point. In this paper

we present an innovative design for the operation of such

ad-hoc networks. The basic idea of the design is to op-

erate each Mobile Host as a sp ecialized router, which

perio dically advertises its view of the interconnection

topology with other Mobile Hosts within the network.

This amounts to a new sort of routing proto col. We

haveinvestigated modications to the basic Bellman-

Ford routing mechanisms, as specied by RIP 5], to

make it suitable for a dynamic and self-starting network

mechanism as is required by users wishing to utilize ad-

hoc networks. Our mo dications address some of the

previous ob jections to the use of Bellman-Ford, related

to the p oor lo oping properties of such algorithms in the

face of broken links and the resulting time dep endent

nature of the interconnection top ology describing the

links b etween the Mobile Hosts. Finally,we describe

the ways in which the basic network-layer routing can

be modied to provide MAC-layer support for ad-ho c

networks.

1 Intro duction

Recently, there has b een tremendous growth in the

sales of laptop and portable computers. These smaller

computers, nevertheless, can b e equipp ed with hun-

dreds of megabytes of disk storage, high resolution color

displays, p ointing devices, and wireless communications

adapters. Moreover, since many of these small (in size

only) computers op erate for hours with battery p ower,

users are free to move about at their convenience with-

out being constrained by wires.

This is a revolutionary development in personal com-

puting. Battery p owered, untethered computers are

likely to become a p ervasive part of our computing in-

frastructure. As people begin to have mobile computers

handy, for whatever purposes, sharing information be-

tween the computers will become a natural requirement.

Currently, such sharing is made dicult by the need for

users to p erform administrative tasks and set up static,

bidirectional links b etween their computers. However,

if the wireless communications systems in the mobile

computers supp ort a broadcast mechanism, much more

exible and useful ways of sharing information can be

imagined. For instance, anynumber of people could

conceivably enter a conference ro om and agree to sup-

port communications links b etween themselves, with-

out necessarily engaging the services of any pre-existing

equipment in the room (i.e, without requiring anypre-

existing communications infrastructure). Thus, one of

our primary motivations is to allow the construction of

temporary networks with no wires and no administra-

tiveintervention required. In this paper, suchainter-

connection b etween the mobile computers will b e called

ad-hoc

network, in conformance with current usage

within the IEEE 802.11 sub committee 4].

Ad-hoc networks dier signicantly from existing

networks. First of all, the top ology of interconnections

may b e quite dynamic. Secondly, most users will not

wish to p erform any administrative actions to set up

suchanetwork. In order to provide service in the most

general situation, we do not assume that every computer

is within communication range of every other computer.

This lack of complete connectivitywould certainly b e a

reasonable characteristic of, say, a population of mo-

bile computers in a large room which relied on infrared

transceivers to eect their data communications.

Currently, there is no method available which enables

mobile computers with wireless data communications

equipment to freely roam ab out while still maintaining

connections with each other, unless special assumptions

are made ab out the way the computers are situated with

respect to each other. One mobile computer mayoften

be able to exchange data with two other mobile comput-

ers which cannot themselves directly exchange data. As

a result, computer users in a conference room maybe

unable to predict which of their associates' computers

could be relied upon to maintain network connection,

especially as the users moved from place to place within

the room.

Routing proto cols for existing networks havenot

been designed specically to provide the kind of dy-

namic, self-starting behavior needed for ad-hoc net-

works. Most protocols exhibit their least desirable be-

havior when presented with a highly dynamic inter-

connection top ology. Although we thought that mo-

bile computers could naturally be modeled as

routers

it was also clear that existing routing proto cols would

place too heavy a computational burden on each mobile

computer. Moreover, the convergence characteristics of

existing routing protocols did not seem goo d enough

to t the needs of ad-ho c networks. Lastly, the wire-

less medium diers in importantways from wired me-

dia, whichwould require that wemake modications to

whichever routing proto col we mightchoose to exper-

iment with. For instance, mobile computers maywell

have only a single network interface adapter, whereas

most existing routers havenetwork interfaces to connect

two separate networks together. Besides, wireless media

are of limited and variable range, in distinction to exist-

ing wired media. Since we had to makelots ofchanges

anyway,we decided to follow our ad-hoc network model

as far as we could and ended up with a substantially

new approach to the classic distance-vector routing.

2 Overview of Routing Methods

In our environment, the problem of routing is essen-

tially the distributed version of the shortest path prob-

lem 11]. Eachnodeinthenetwork maintains for each

destination a preferred neighbor. Each data packet con-

tains a destination node identier in its header. When

a no de receives a data packet, it forwards the packet

to the preferred neighbor for its destination. The for-

warding process continues until the packet reaches its

destination. The manner in which routing tables are

constructed, maintained and updated diers from one

routing metho d to another. Popular routing methods,

however, attempt to achieve the common ob jectiveof

routing packets along the optimal path. The next-hop

routing metho ds can be categorized into two primary

classes:

link-state

and

distance-vector

Link-State

The link-state approach is closer to the

centralized version of the shortest path computation

method. Eachnodemaintains a view of the network

topology with a cost for eachlink. Tokeep these views

consistent, each no de periodically broadcasts the link

costs of its outgoing links to all other nodes using a

protocol such as ooding. As a node receives this in-

formation, it updates its view of the network topology

and applies a shortest-path algorithm to choose its next

hop for each destination. Some of the link costs in a

node's view can be incorrect b ecause of long propaga-

tion delays, partitioned network, etc. Such inconsistent

views of network topologies might lead to formation of

routing loops. These loops, however, are short-lived,

because they disappear in the time it takes a message

to traverse the diameter of the network 8].

Distance-Vector

In distance-vector algorithms,

every node

maintains, for each destination

, a set

of distances

where

ranges over the neighbors of

.Node

treats neighbor

as a next-hop for a packet

destined for

equals

min

. The succession

of next hops chosen in this manner lead to

along the

shorest path. In order to keep the distance estimates

up-to-date, each no de monitors the cost of its outgo-

ing links and periodically broadcasts, to each one its

neighbors, its current estimate of the shortest distance

to every other node in the network.

The above distance-vector algorithm is the classical

Distributed Bellman-Ford (DBF) algorithm 2]. Com-

pared to link-state metho d, it is computationally more

ecient, easier to implement and requires muchless

storage space. However, it is well known that this algo-

rithm can cause the formation of both short-lived and

long-lived lo ops 3]. The primary cause for formation

of routing lo ops is that nodes choose their next-hops in

a completely distributed fashion based on information

which can p ossibly be stale and, therefore, incorrect.

Almost all proposed modications to DBF algorithm

6, 7, 9] eliminate the looping problem by forcing all

nodes in the network to participate in some form of in-

ternodal co ordination proto col. Suchinternodal coordi-

nation mechanisms might b e eective when top ological

changes are rare. However, within an ad-ho c mobile en-

vironment enforcing anysuchinternodal coordination

mechanism will b e dicult due to the rapidly changing

topology of the underlying routing network.

Simplicity is one of the primary attributes which

makes any routing protocol

preferred

over others for

implementation within operational networks. RIP 5]

is a classical example. Despite the

counting-to-innity

problem it has proven to be very successful within small

size internetworks. The usefulness of RIP within ad-ho c

environment, however, is limited as it was not designed

to handle rapid topological changes. Furthermore, the

techniques of

split-horizon

and

poisoned-reverse

5] are

not useful within the wireless environment due to the

broadcast nature of the transmission medium. Our de-

sign goal therefore has b een to design a routing metho d

for ad-hoc networks whichpreserves the simplicityof

RIP,yet at the same time avoids the looping problem.

Our approachistotageach route table entry with a

sequence number so that no des can quickly distinguish

stale routes from the new ones and thus avoid formation

of routing lo ops.

3 Destination-Sequenced Distance Vec-

tor (DSDV) Protocol

Our proposed routing method allows a collection of

mobile computers, whichmay not be close to anybase

station and can exchange data along changing and arbi-

trary paths of interconnection, to aord all computers

among their number a (possibly multi-hop) path along

which data can b e exchanged. In addition, our solution

must remain compatible with operation in cases where

a base station is available. By the methods outlined in

this pap er, not only will routing be seen to solvethe

problems associated with ad-hoc networks, but in ad-

dition we will describe ways to perform such routing

functions at Layer 2, which traditionally has not been

utilized as a proto col level for routing.

Packets are transmitted b etween the stations of the

network by using routing tables which are stored at

each station of the network. Each routing table, at each

of the stations, lists all available destinations, and the

number of hops to each. Each route table entry is tagged

with a sequence number which is originated by the des-

tination station. To maintain the consistency of routing

tables in a dynamically varying top ology,each station

perio dically transmits updates, and transmits updates

immediately when signicant new information is avail-

able. Since we do not assume that the mobile hosts

are maintaining anysort oftimesynchronization, we

also make no assumption about the phase relationship

of the update perio ds b etween the mobile hosts. These

packets indicate which stations are accessible from each

station and the number of hops necessary to reach these

accessible stations, as is often done in distance-vector

routing algorithms. It is not the purp ose of this paper to

propose any new metrics for route selection other than

the freshness of the sequence numbers asso ciated with

the route cost or other metrics might easily replace the

number of hops in other implementations. The packets

may be transmitted containing either layer 2 (MAC)

addresses or layer 3 (network) addresses.

Routing information is advertised by broadcasting or

multicasting the packets which are transmitted p eriod-

ically and incrementally as top ological changes are de-

tected { for instance, when stations movewithinthe

network. Data is also kept about the length of time be-

tween arrival of the

rst

and the arrival of the

best

route

for each particular destination. Based on this data, a

decision may be made to delayadvertising routes which

are about to change soon, thus damping uctuations

of the route tables. The advertisementofrouteswhich

maynothave stabilized yet is delayed in order to reduce

the number of rebroadcasts of possible route entries that

normally arrive with the same sequence number.

The DSDV proto col requires each mobile station to

advertise, to each of its current neighbors, its own rout-

ing table (for instance, by broadcasting its entries). The

entries in this list maychange fairly dynamically over

time, so the advertisementmust be made often enough

to ensure that every mobile computer can almost al-

ways lo cate every other mobile computer of the collec-

tion. In addition, each mobile computer agrees to relay

data packets to other computers upon request. This

agreement places a premium on the ability to deter-

mine the shortest number of hops for a route to a desti-

nation wewould liketo avoid unnecessarily disturbing

mobile hosts if they are in sleep mo de. In this wayamo-

bile computer may exchange data with any other mobile

computer in the group even if the target of the data is

not within range for direct communication. If the noti-

cation of which other mobile computers are accessible

from any particular computer in the collection is done

at layer 2, then DSDVwillwork with whatever higher

layer (e.g., Network Layer) protocol might be in use.

All the computers interoperating to create data paths

between themselves broadcast the necessary data pe-

riodically,say once every few seconds. In a wireless

medium, it is imp ortanttokeep in mind that broad-

casts are limited in range bythephysical characteris-

tics of the medium. This is dierent than the situation

with wired media, which usually haveamuchmorewell-

dened range of reception. The data broadcast byeach

mobile computer will contain its new sequence number

and the following information for each new route:



The destination's address



The number of hops required to reach the destina-

tion and



The sequence number of the information received

regarding that destination, as originally stamped

by the destination

The transmitted routing tables will also contain the

hardware address, and (if appropriate) the network ad-

dress, of the mobile computer transmitting them, within

the headers of the packet. The routing table will also

include a sequence number created by the transmitter.

Routes with more recent sequence numbers are always

preferred as the basis for making forwarding decisions,

but not necessarily advertised. Of the paths with the

same sequence number, those with the smallest metric

will be used. By the natural way in which the routing

tables are propagated, the sequence number is senttoall

mobile computers whichmayeach decide to maintain a

routing entry for that originating mobile computer.

Routes received in broadcasts are also advertised by

the receiver when it subsequently broadcasts its routing

information the receiver adds an increment to the met-

ric before advertising the route, since incoming pack-

ets will require one more hop to reach the destination

(namely, the hop from the transmitter to the receiver).

Again, we do not explicitly consider here the changes

required to use metrics which do not use the hop count

to the destination.

One of the most imp ortant parameters to b e chosen is

the time b etween broadcasting the routing information

packets. However, when any new or substantially modi-

ed route information is received by a Mobile Host, the

new information will be retransmitted so on (sub ject to

constraints imposed for damping route uctuations), ef-

fecting the most rapid possible dissemination of routing

information among all the coop erating Mobile Hosts.

This quick re-broadcast introduces a new requirement

for our protocols to converge as soon as possible. It

would be calamitous if the movement of a Mobile Host

caused a storm of broadcasts, degrading the availability

of the wireless medium.

Mobile Hosts cause broken links as they move from

place to place. The broken link may b e detected by

the layer-2 protocol, or it may instead b e inferred if no

broadcasts have been received for a while from a for-

mer neighbor. A broken link is described by a metric of

(i.e., anyvalue greater than the maximum allowed

metric). When a link to a next hop has broken, any

route through that next hop is immediately assigned an

metric and assigned an updated sequence number.

Since this qualies as a substantial route change, such

modied routes are immediately disclosed in a broad-

cast routing information packet. Building information

to describ e broken links is the only situation when the

sequence numberisgeneratedbyany Mobile Host other

than the destination Mobile Host. Sequence numbers

dened by the originating Mobile Hosts are dened to

be even numbers, and sequence numbers generated to

indicate

metrics are o dd numbers. In this wayany

"real" sequence numbers will sup ersede an

metric.

When a no de receives an

metric, and it has a later

sequence number with a nite metric, it triggers a route

update broadcast to disseminate the importantnews

about that destination.

In a very large population of Mobile Hosts, adjust-

ments will likely b e made in the time between broad-

casts of the routing information packets. In order to re-

duce the amount of information carried in these packets,

twotypes will be dened. One will carry all the avail-

able routing information, called a "full dump". The

other type will carry only information changed since

the last full dump, called an "incremental". By design,

an incremental routing up date should t in one net-

work protocol data unit (NPDU). The full dump will

most likely require multiple NPDUs, even for relatively

small p opulations of Mobile Hosts. Full dumps can be

transmitted relatively infrequently when no movement

of Mobile Hosts is occurring. When movement b ecomes

frequent, and the size of an incremental approaches the

size of a NPDU, then a full dump can b e scheduled (so

that the next incremental will b e smaller). It is expected

that mobile no des will implement some means for de-

termining which route changes are signicant enough to

be sent out with each incremental advertisement. For

instance, when a stabilized route shows a dierentmet-

ric for some destination, that would likely constitute

a signicantchange that needed to b e advertised after

stabilization. If a new sequence numberforarouteis

received, but the metric stays the same, that would b e

unlikely to b e considered as a signicantchange.

When a Mobile Host receives new routing informa-

tion (usually in an incremental packet as just describ ed),

that information is compared to the information al-

ready available from previous routing information pack-

ets. Any route with a more recent sequence number

is used. Routes with older sequence numbers are dis-

carded. A route with a sequence number equal to an

existing route is chosen if it has a "b etter" metric, and

the existing route discarded, or stored as less prefer-

able. The metrics for routes chosen from the newly

received broadcast information are each incremented

by one hop. Newly recorded routes are scheduled for

immediate advertisement to the current Mobile Host's

neighbors. Routes whichshowanimproved metric are

scheduled for advertisement at a time which dep ends

on the average settling time for routes to the particular

destination under consideration.

Timing skews b etween the various Mobile Hosts are

expected. The broadcasts of routing information bythe

Mobile Hosts are to b e regarded as somewhat asyn-

chronous events, even though some regularityisex-

pected. In such a population of indep endently trans-

mitting agents, some uctuation could develop using the

above pro cedures for updating routes. It could turn out

that a particular Mobile Host would receive new routing

information in a pattern which causes it to consistently

change routes from one next hop to another, even when

the destination Mobile Host has not moved. This hap-

pens because there are twoways for new routes to b e

chosen they mighthave a later sequence number, or

they mighthave a b etter metric. A Mobile Host could

conceivably always receivetwo routes to the same des-

tination, with a newer sequence number, one after an-

other (via dierent neighbors), but always get the route

with the worse metric rst. Unless care is taken, this

will lead to a continuing burst of new route transmittals

upon every new sequence number from that destination.

Each new metric is propagated to every Mobile Host in

the neighborho od, which propagates to their neighbors

and so on.

One solution is to delaytheadvertisementofsuch

routes, when a Mobile Host can determine that a route

with a better metric is likely to showupsoon. The

route with the later sequence number must be available

for use, but it does not havetobeadvertised imme-

diately unless it is a route to a destination whichwas

previously unreachable. Thus, there will be tworout-

ing tables keptateach Mobile Host one for use with

forwarding packets, and another to be advertised via

incremental routing information packets. To determine

the probability of imminentarrival of routing informa-

tion showing a b etter metric, the Mobile Host has to

keep a history of the weighted average time that routes

to a particular destination uctuate until the route with

the best metric is received. We hope that suchapro-

cedure will allow us to predict howlongtowait before

advertising new routes.

Operating DSDVatLayer 2

The addresses stored in the routing tables will cor-

respond to the layer at which this ad-hoc networking

protocol is operated. That is, op eration at Layer 3 will

use network layer addresses for the next hop and des-

tination addresses, and op eration at Layer 2 will use

Layer 2 Media Access Control (MAC) addresses.

Using MAC addresses for the forwarding table does

introduce a new requirement, however. The diculty

is that Layer 3 network protocols provide communica-

tion based on network addresses, and a waymust be

provided to resolvethese Layer 3 addresses into MAC

addresses. Otherwise, a multiplicity of dierent address

resolution mechanisms would b e put into place, and a

corresponding loss of bandwidth in the wireless medium

would b e observed whenever the resolution mechanisms

were utilized. This could be substantial since such

mechanisms would require broadcasts and retransmit-

ted broadcasts byevery Mobile Host in the ad-hoc net-

work. Thus, unless special care is taken, every address

resolution mightlook like a glitch in the normal op er-

ation of the network, whichmaywell be noticeable to

any active users.

The solution prop osed here, for op eration at Layer 2,

is to include Layer 3 proto col information along with the

Layer 2 information. Each destination host would ad-

vertise whichLayer 3 protocols it supports, and each

Mobile Host advertising reachability to that destina-

tion would include along, with the advertisement, the

information about the Layer 3 protocols supp orted at

that destination. This information would only haveto

be transmitted when it changes, which occurs rarely.

Changes would b e transmitted as part of each incremen-

tal dump. Since each Mobile Host could supp ort several

Layer 3 protocols (and many will), this list would have

to be variable in length.

Extending Base Station Coverage

Mobile computers will frequently be used in conjunc-

tion with base stations, whichallowthemtoexchange

data with other computers connected to the wired net-

work. By participating in the DSDV protocol, base

stations can extend their coverage beyond the range

imposed by their wireless transmitters. When a base

station participates in DSDV, it is shown as a default

route in the tables transmitted by a mobile station. In

this way, mobile stations within range of a base station

can cooperate to eectively extend the range of the base

station to serve other stations outside the range of the

base station, as long as those other mobile stations are

close to some other mobile station that is within range.

4 Examples of DSDV in operation

Figure 1: Movement in an ad-hoc network

Consider

in Figure 1. Table 1 shows a p ossible

structure of the forwarding table whichis maintained

. Suppose the address

.ofeach Mobile Host is

represented as

Suppose further that all sequence

numbers are denoted SNNN

, where

spec-

ies the computer that created the sequence number

and SNNN is a sequence number value. Also supp ose

that there are entries for all other Mobile Hosts, with se-

quence numbers SNNN

, b efore

moves away

from

. The install time eld helps determine when

If DSDV is operated at level 2 then

denotes the MAC

address, otherwise it denotes a level 3 address

Highly dynamic Destination-Sequenced Distance-Vector routing (DSDV) for mobile computers

Citations

Dynamic Source Routing in Ad Hoc Wireless Networks

GPSR: greedy perimeter stateless routing for wireless networks

A performance comparison of multi-hop wireless ad hoc network routing protocols

Epidemic routing for partially-connected ad hoc networks

A review of current routing protocols for ad hoc mobile wireless networks

References

Routing Information Protocol

The New Routing Algorithm for the ARPANET

Routing Techniques Used in Computer Communication Networks

A unified approach to loop-free routing using distance vectors or link states

A Failsafe Distributed Routing Protocol

Related Papers (5)

Ad-hoc on-demand distance vector routing

Dynamic Source Routing in Ad Hoc Wireless Networks

Ad hoc On-Demand Distance Vector (AODV) Routing

A performance comparison of multi-hop wireless ad hoc network routing protocols

A review of current routing protocols for ad hoc mobile wireless networks

Frequently Asked Questions (7)

Q1. What contributions have the authors mentioned in the paper "Highly dynamic destination sequenced distance vector routing dsdv for mobile computers" ?

Q2. what is the way to determine if a route is a neighbor?

Q3. What is the reason why the loop free property holds in the second scenario?

Q4. What is the primary cause of loop formation in BF style algorithms?

Q5. how many different structures are used to represent the different types of routes?

Q6. What is the way to get rid of the uctuation problem?

Q7. What is the main concern of ad hoc routing?