I Pv 6 Autoconfiguration in Large Scale Mobile Ad-Hoc Networks ...

steambeanSoftware and s/w Development

Jun 30, 2012 (4 years and 11 months ago)


IPv6 Autoconfiguration in Large Scale Mobile Ad-Hoc Networks
Kilian Weniger

,Martina Zitterbart
Institute of Telematics
University of Karlsruhe,76128 Karlsruhe,Germany
Ph:++49 721 608-{6415/6400},Email:{weniger,zit}@tm.uka.de
Mobile ad-hoc networks are infrastructure-free,
highly dynamic wireless networks,where central ad-
ministration or configuration by the user is impractical.
The Internet Protocol IPv6 defines mechanisms to auto-
configure interfaces of nodes in wired networks in a dis-
tributed manner.In this paper,the applicability of IPv6
Stateless Address Autoconfiguration and IPv6 Neighbor
Discovery Protocol to large scale mobile ad-hoc net-
works is investigated.A hierarchical approach based on
so-called leader nodes is proposed together with a leader
election algorithm.Address autoconfiguration with IPv6
in very dynamic ad-hoc networks requires special sup-
port as,for example,outlined within this paper.Exten-
sions to the IPv6 Neighbor Discovery Protocol are pro-
posed to enable an efficient and scalable usage in ad-hoc
In mobile ad-hoc networks,nodes spontaneously form
a network,where each node can reach not only nodes in
its direct transmission range,but also distant nodes,ac-
cessible through a chain of intermediate nodes.Figure
1 illustrates a scenario,where node n
sends a packet
over multiple hops towards its journey to the destination
.Because of node mobility,the available paths to a
destination change over time with a considerably higher
frequency than in infrastructure networks.Furthermore,
the network may partition and merge again later,depen-
dent on the mobility patterns and transmission ranges of
the mobile nodes.Consequently,enhanced requirements
exist for routing protocols operating in such an environ-
ment.This has been addressed by various research ef-
forts on mobile ad-hoc routing (e.g.,Monarch Project
at CMU [1],Wireless Adaptive Mobility Laboratory at
UCLA [2]).
Much less attention has been given to autoconfigura-
tion mechanisms,such as address autoconfiguration.In
general,the purpose of address autoconfiguration is the
assignment of an address to an interface,which is unique
and routable in the network.In ad hoc networks,this
mechanism need to cope with the high dynamics within
such networking environments.
This work was supported by the German Federal Ministry of Ed-
ucation and Research (BMBF).It was part of the IPonAir project be-
longing to the research focus hyperNET.HyperNET stands for Univer-
sal Utilization of Communications Networks for Future Generations of
Mobile Communications Systems.
In this paper,we investigate the applicability of IPv6
autoconfiguration mechanisms to ad hoc networking.It is
assumed,that the mobile nodes will be identified through
IPv6 addresses.Figure 1:Multi-hop ad-hoc Routing
Transmission Range
The approach presented in this paper is based on the
following requirements:
 The mechanism needs to cope with the network dy-
namics present in mobile ad-hoc networks.There-
fore,a distributed approach is used.
 The mechanismshall scale to large ad-hoc networks
with,e.g.,thousands of nodes with multiple inter-
faces.Among others,address space limitations and
signaling traffic load must be considered carefully.
 Because there exist many routing protocols for ad
hoc networks,each optimized for special network
scenarios,the mechanism shall be independent of
the routing protocol.However,optimizations are
possible if both are cooperating.
 If two independently configured ad hoc networks
merge,the uniqueness of the addresses shall be guar-
anteed afterwards,i.e.network partitioning and
merging shall be supported.
Because IPv6 is in discussion for mobile communi-
cation systems beyond the 3rd generation,we adopt the
IPv6 Stateless Address Autoconfiguration (SAA) [3] and
the corresponding Neighbor Discovery Protocol (NDP)
[4] [5] to ad-hoc networks with the goal of scalability in
mind.We choose a hierarchical approach,where some
nodes are responsible for parts of the address configura-
tion of other nodes.
The paper is structured as follows:Related research ef-
forts are discussed in section 2.A brief overview of IPv6
Stateless Address Autoconfiguration is given in section
3.Section 4 describes howto adapt the process to ad-hoc
networks and presents an election algorithmfor the leader
nodes.Section 5 discusses the interaction with routing
protocols.And,finally,section 6 concludes the paper.
Within the IETF working group for Mobile Ad-Hoc
Networks (MANET) several routing protocols were pro-
posed.These protocols can be roughly classified into
proactive and reactive protocols [6] [7] [8].Most proac-
tive protocols evolved from distance-vector or link-state
routing protocols.By periodically exchanging control
messages with neighboring nodes,each node maintains
an up-to-date version of the network topology.In con-
trast,reactive protocols,e.g.the Dynamic Source Rout-
ing protocol (DSR) [9],discover the route to the destina-
tion on demand.Depending on network dynamics,size,
topology and traffic patterns,each protocol has its advan-
tages and drawbacks.Hybrid protocols try to combine
both approaches,mostly by building a hierarchy.
Address autoconfiguration in large ad-hoc networks is
still an unresolved issue.Central administration,e.g.,
based on DHCP,or manual configuration by the user is
impossible in large mobile ad-hoc networks.
The IETF ZEROCONF working group deals with au-
toconfiguration issues,but with a focus on wired net-
works.Asimple solution for address autoconfiguration in
ad-hoc networks by using IPv4 link-local addresses was
proposed by Perkins [10].Addresses are randomly cho-
sen from the network 169.254/16.Duplicate addresses
are detected by using protocol messages of the route dis-
covery phase of a reactive routing protocol like DSR [9].
Duplicate Address Detection (DAD) is performed only
once by each node.Subsequently,the uniqueness of
addresses is not guaranteed in partitioned networks that
merge later on.Each node performing the DAD uses
the route discovery mechanism of the routing protocol.
This mechanism,in turn,is based on flooding the en-
tire network to find the destination.This is not suitable
for large scale ad-hoc networks,especially if the DAD is
repeated to support network merging.Furthermore,the
address space is limited to 2047 during and 65535 af-
ter the DAD.The probability of an address conflict in-
creases with the number of interfaces that wish to oper-
ate in the network.It should be considered that mobile
nodes may have multiple interfaces to different ad hoc
networks.During DAD,a tentative IPv4 address is used
as source address,which is selected randomly out of the
range 1-2047.Although these addresses should never be
used for more than a few seconds,problems may arise in
large ad hoc networks,especially if a lot of nodes power
on at the same time.In summary,the approach presented
in [10] only works with specific proactive routing proto-
cols,has address space limitations and does not cope with
dynamic network partitioning and merging.
Park et.al.[11] propose to use site-local IPv6 ad-
dresses and Neighbor Discovery for DAD.The use of
IPv6 addresses reduces the problem of the limited ad-
dress space,but leaves a lot of issues unresolved.In [11],
subnet IDs are selected randomly by each node and are
not used to divide the network into subnetworks.This
results in a flat address scheme.Usually the subnet IDs
are assigned by routers on the link using Router Adver-
tisements (RA).In summary,this approach requires all
nodes to flood the entire ad-hoc network with forwarded
NDP messages and does not consider network merging
as well.
Another approach [12] [13],called the Dynamic Reg-
istration and Configuration Protocol (DRCP),tries to ex-
tend DHCP to a stateless autoconfiguration protocol for
wired and wireless networks.Therefore,each node rep-
resents a DRCP client and server and owns an IPv4 ad-
dress pool.The Dynamic Address Allocation Protocol
(DAAP) is responsible for the distribution of the address
pools.Therefore,each node requesting a pool obtains
half of the pool of a neighboring node.This may lead to
a lot of unassigned addresses in the already scarce IPv4
private network address space and,subsequently,to scal-
ability problems.Network merging is not considered as
The IPv6 SAA[3] is based on NDP [4] which is speci-
fied for links that support a native form of multicast or
broadcast.Some other links are covered by extension
documents,but there is no extension for ad hoc networks
The IPv6 SAA basically consists of three phases:
1.Construction of a link-local address for the use on
the local link
2.Duplicate Address Detection
3.Construction of a site-local address for the use on
the site
The use of a global address is not part of the SAA.There-
fore,Stateful Address Autoconfiguration (DHCPv6) [14]
may be used.
The process begins with the construction of a link-
local address that is based on the interface identifier and
a well-known link-local prefix.It is first considered as a
tentative address.IEEE defines a 64-bit Extended Uni-
versal Identifier (EUI-64),which is used for this pur-
pose.It is calculated,e.g.,from the MAC-address of an
IEEE 802.x interface.Although the EUI-64 number is
designed to be globally unique,this cannot be guaran-
teed,because of manufacturers using unregistered 802.x-
addresses,support for changing the MAC-address by the
user or obscure network devices,which choose a MAC-
address randomly.The DAD process is needed to detect
and handle duplicate addresses.Therefore,the node is-
sues a Neighbor Solicitation (NS) message (see Figure 2)
with the well-known unspecified address as source IP ad-
dress.This address must never be assigned to a node and
indicates the absence of an address [15].Figure 2:The Neighbor Solicitation (NS) message
Payload Length Next Header
Type: 135 Code: 0
Target Address:
tentative address
Destination Address:
solicited−node multicast address
Source Address:
IPv6 Header
Flow Label
unspecified address
NS message
Hop Limit: 255
The solicited-node multicast address is used as desti-
nation IP address,which is based on a well-known prefix
and the last 24 bits of the tentative address.The tenta-
tive address is used as solicitation target address and a
hop limit of 255 is used to limit the message on the local
link.NDP messages with a hop limit of less than 255 are
discarded.This prevents hackers from introducing NDP
messages to the local link fromoutside.
If the address is already in use by another node,this
node responds with a Neighbor Advertisement (NA) mes-
sage carrying the all-nodes multicast address as destina-
tion IP address.An address conflict is recognized,if the
sender receives a NA message in reply to the NS mes-
sage or if a NS message with the same solicitation target
address is received,indicating that another node with the
same tentative address currently performs the DAD.
If Router Advertisements containing a subnet ID are
received,hosts construct a site-local address using the
link-local address,a well-known site-local prefix and the
announced subnet ID.
4.1 Hierarchical Approach
Nodes in mobile ad-hoc networks are integrated hosts
and routers.As a consequence,all nodes execute all
router functionalities.The pre-requisite to apply IPv6
SAA in environments other than a single broadcast link
is the presence of routers,which issue Router Advertise-
ments (RA).During the DAD,flooding is needed,be-
cause all routing or multipoint relaying algorithms as-
sume unique node identifiers.The flooding leads to the
scalability problems already mentioned.In order to limit
the flooding to a bounded area,broadcast links"emu-
lated".Each node defines its broadcast link,from now
on called scope,as the group of nodes,that are less or
equal than r
hops away.A hierarchy is established by
special nodes,the so-called leader nodes,that configure
a group of nodes by issuing Router Advertisements (RA).
The routing protocol used in the ad-hoc network does
not necessarily need to follow this hierarchical structure.
However,it may be advantageous if it does (see section
5).Furthermore,it is preferable that nodes move in logi-
cal groups.Otherwise,the cost of maintaining the hierar-
chical structure may increase considerably.
4.2 Link-Local Address and Duplicate Address
A mobile node joining the ad-hoc network first gener-
ates a tentative link-local address.Subsequently,DAD
needs to be performed.In order to do so,the mobile
node issues a modified NS message containing the ten-
tative link-local address and a hop limit of r
.The lat-
ter limits the messages to the scope of the node.Conse-
quently,the scope forms an abstraction of a broadcast link
in wired LANs and hides the multi-hop structure from
the SAA process.Furthermore,nodes use the all-nodes
multicast address instead of the solicited-node multicast
address.Packets sent to the all-nodes multicast address
are received and processed by all nodes.A multicast tree
cannot be established,because the uniqueness of the IP
addresses cannot be assumed.When a NDP message is
received,the nodes decrement the hop limit field and for-
ward the message.Forwarding only takes place,if the
hop limit is greater than zero.
The detection of an address conflict used in fixed net-
works cannot be directly applied to ad-hoc networks,be-
cause each mobile node receives echos of recently sent
NS and NA messages multiple times,depending on the
number of adjacent nodes receiving and forwarding this
message.This number,however,is not known to the mo-
bile host and changes over time.Additionally,a message
sent by a node cannot be distinguished from a message
originated from another node,that performs DAD and
that has the same tentative address.This means,that an
address conflict cannot be detected.The messages are
completely identical:The source address is the unspeci-
fied address,the destination address is the solicited-node
multicast address and the target address is the tentative
address.Therefore,we propose to define a new option
for Neighbor Discovery messages,called the Mobile Ad-
Hoc Networks (MANET) option,which contains a Ran-
dom Source ID (RS-ID) field (see Figure 3).It’s value is
randomly chosen by a mobile node for each NDP mes-
sage sent and enables the distinction of messages sent by
different nodes.
Nodes cache the RS-IDof each message received for a
certain period of time and only forward NDP messages
with RS-IDs that are not in the cache.If a scope is
densely populated,a broadcast storm resulting from all
the forwarded messages can lead to heavy contention and
collisions.This problem is alleviated by,e.g.,a counter-
based scheme as proposed in [16].In this case,a message
is forwarded only,if the node does not receive the same
message forwarded by adjacent nodes more than a pre-
defined number of times.
If the DAD succeeds,the address can be considered as
valid for a certain period of time.With the mechanism
outlined above,link-local addresses are guaranteed to be
unique within the scope of each node.But they are not
yet guaranteed to be unique within the entire ad-hoc net-
work.If the DAD fails,manual configuration is needed
or another tentative address may be chosen randomly.
In summary,the following approaches are applied to
overcome shortcomings of NDP with respect to ad-hoc
 At first,NS messages are limited to a single hop.
In order to extend NDP for multi-hop use,a hop
limit of r
instead of 255 is used and the solicited-
node address as destination IP address is replaced by
the all-nodes multicast address in NS and NA mes-
sages.Furthermore,NDP messages with a hop limit
less than r
and bigger than 0 are forwarded within
the scope using a counter-based scheme to prevent
broadcast storms.Because the messages are lim-
ited to the scope,the mechanismworks even in large
scale ad-hoc networks.
 The detection of duplicate addresses cannot be ap-
plied in a multi-hop environment as described in the
specification.Therefore,the RS-ID is introduced to
distinguish NS and NAmessages sent by nodes dur-
ing the DAD.Figure 3:The modified Neighbor Solicitation message
Payload Length Next Header
Type: 135 Code: 0
Target Address:
tentative address
Destination Address:
all−nodes multicast address
Source Address:
IPv6 Header
Version Flow Label
unspecified address
Hop Limit: r
Type: 6 Length: 2 Random Source ID
Isolated Node Flag
# Nodes in ScopeNode Status
MANET option
NS message
Prerelay Hop L.
with the MANET option
4.3 Leader election algorithm
In order to establish the hierarchical structure and guar-
antee network wide unique addresses,leader nodes need
to be elected.The election algorithmshould followsome
requirements to function in a proper and efficient manner:
 The number of leader nodes in the entire ad-hoc net-
work should be small compared to the overall num-
ber of nodes.This assures an efficient operation in
terms of signaling traffic overhead and an economic
handling of the limited number of available subnet
 Leader nodes should not change too frequent in or-
der to keep the network topology stable and to keep
the signaling overhead as small as possible.
 Unique node identifiers cannot be assumed,al-
though it is assumed that duplicate identifiers are
The first requirement is accommodated by choosing
the node with the highest number of neighbors as the
leader node.In order to determine this number,the elec-
tion algorithmcan benefit fromNS messages sent during
the DAD procedure.
In order to keep changes of leader nodes infrequent,
the following states are introduced for a node:
 Leader state (S),
 Candidate state (C),or
 Host state (H).
The leader node is in leader state.One or more nodes,
that are not in leader state,but have the highest number
of neighbors in the scope,qualify themselves as potential
leader nodes and,thus,are in candidate state.All other
nodes are in host state.The number of neighbors is de-
termined by counting the entries in the Neighbor Cache.
Therefore,solicitation target addresses of exchanged NS
messages need to be saved in the Neighbor Cache for a
limited time.These entries are marked with a special flag.
Nodes learn the number of neighbors n
from other
nodes in the scope by exchanging the modified Neigh-
bor Solicitation containing a Number of Nodes in Scope
field during DAD.In order to maintain the hierarchy de-
spite of node mobility,the leader election needs to be per-
formed periodically by exchanging NS messages within
the scope.Simultaneously,unique addresses can be guar-
anteed,even in the case of network partitioning and merg-
ing.The period for the DAD is t
,which is assumed
to be equal for all nodes in the network.A Node Status
field (see Figure 3) indicates,whether this message was
sent by an actual leader node.
Furthermore,the election algorithmcan be subdivided
into three tasks:
1.Receiving and processing NDP messages:
The number of neighbors is determined based on the
number of received NS messages.If an NS mes-
sage is received from a leader node,the number of
neighbors n
of the sender is stored in n
quently,each node knows the number of neighbors
of the current leader node.Otherwise,the number is
compared to the number of the current candidate n
If n
> n
,the sender of the NS message is a new
candidate and n
is replaced by n
is left unchanged and the sender is in host state.The
leader node is not affected by this procedure.
2.Changing the leader:
If the DAD timer expires,the number of neighbors
of the corresponding node is compared to the num-
ber of neighbors of the leader node and the candidate
node.If the node is in leader state and n < n
the node withdraws its leader state.Otherwise,if
n is equal to n
,the node is a candidate node.If
n > n
+ n
,the candidate nodes have significant
more neighbors than the actual leader node.The
one with the highest RS-ID becomes the new leader
node.To prevent oscillations of leader nodes,a hys-
teresis represented by n
is used to lead the replace-
ment of a leader node.If the leader node powers off
or leaves the network,n
is zero in the next round
and the node in candidate state becomes the new
leader node.
3.Sending an NS message:
If the DAD timer expires,an NS messages is is-
sued containing the actual number of neighbors and
the state of the node represented by the Node Status
and n
are initialized to zero and a
new period t
The election process outlined above is illustrated in the
flow diagramshown in figure 4.
The optimal choice of the system parameters n
depends on the node movement and the scope size.
They are subject to further.
4.4 Forming a Site-Local Address
The interface that is in leader state,subscribes to the
all-routers multicast group,chooses a subnet ID ran-
domly and generates a site-local address based on the
link-local address and the subnet ID.The leader node
sends a Router Advertisement (RA) messages within its
scope containing the subnet ID and its link-local address
as source address.Only the link-local address may be
used for this purpose to avoid confusion about the sender
of the RA messages.
All nodes within the scope of the leader node receive
the subnet ID and generate their site-local address.Be-
cause the link-local address is only guaranteed to be
unique within the scope of each node and the site-local
address is constructed fromthe link-local address and the
subnet ID,the subnet ID needs to be unique within the
entire ad-hoc network.
To ensure this,a DADhas to be performed between the
leader nodes within the entire ad-hoc network.The NS
message would use the all-routers multicast address as
destination address and the site-local address as solicited
target address.Because more than 255 hops shall be sup-
ported,a hop limit of 255 is defined as unlimited for-
warding in this context.Furthermore,the RS-IDs of these
messages need to be cached for a much longer time than
for the DADwithin the scope,because the time a message
needs to travel the entire ad hoc network is much higher.
Flooding of the entire network cannot be avoided here,
but it is limited.Only leader nodes need to flood the net-
work and their number is small compared to the overall
number of nodes in the network.Because the uniquenessFigure 4:The leader election algorithm
Init DAD Timer
and Node Status S
new NS received
n > n n = n
n = n
in L state?
n < n − n
L state
1 2
with n neighbors
S = L ?
change to
L state
DAD Timer
n = 0
n = 0
n = n
send NS msg
n > n + n
and C?
of the subnet IDs shall even be guaranteed in the case that
two independent networks merge,the DAD between the
leader nodes needs to be repeated.The easiest way would
be to do that periodically.A better way is to somehow
detect the situation of merging networks.This could be
done with the aid of the Neighbor Cache:If the marked
entries in the Neighbor Cache suddenly increase dramat-
ically and if it was not empty before,it is concluded that
two networks merged.Subsequently,this node triggers
a DAD process between all leader nodes.Neighboring
nodes detecting the same situation recognize this and do
not trigger the DAD again.Additionally,a timer can en-
sure,that the DAD is not repeated too often in the case
the network partitions and merges frequently.Further-
more,a leader node introducing a new subnet ID needs
to performa DAD of its subnet ID.If the DAD fails,one
leader node needs to choose a new subnet ID.
Subsequently,the nodes generate new site-local ad-
dresses and a process called graceful site renumbering
takes place.Therefore,each node is allowed to keep us-
ing the old address for ongoing connections for a special
period of time,the so-called preferred lifetime.After ex-
ceeding the valid lifetime,the node is forbidden to use
the old address at all.
Because the DAD is done within the scope of each
node,which is obviously different from the scope of the
leader node,duplicate site-local addresses may occur.
One of such scenarios is illustrated in figure 5,where the
leader node B is in between node A and C.Despite Node
A and B are bound to the same leader node,they are out-
side the scope of each other and,therefore,they cannot
detect that they own the same link-local address.Addi-
tionally,they build the same site-local address,because
they use the same subnet ID.Asolution is,that the leader
node forwards each NS and NA message that it receives
with the hop limit set to r
.Consequently,the message
sent by node A reaches node C and vice versa.This way,
the scope of each node is extended by the scope of the
leader node and the conflict is resolved.The uniqueness
of the site-local addresses of all nodes in the entire ad-
hoc network is ensured.Because these message should
not be counted for the leader election,a Prerelay Hop
Limit field in the modified NS message is defined,which
contains the value of the hop limit field in the IPv6 header
before manipulation by the relaying node.This ensures
the recognition of messages relayed by the leader node
and the distance to the leader node.Figure 5:Two nodes within the same leader node scope,
Node A
Head Node B
Node C
but outside the scope of each other
If a leader node is replaced by a candidate node,the
new leader node adopts the former subnet ID.Thus,no
renumbering of the site is needed.If a node moves from
one scope of a leader node to another,the graceful renum-
bering is used only for this node.This situation is de-
tected by the absence of RAs with the former subnet ID
and the presence of RAs with a new subnet ID.The node
receives the newsubnet IDand generates a newsite-local
A node may be within the scope of more than one
leader node.In this case,the node is multi-homed and
needs to generate many site-local addresses.A scenario
with two leader nodes is illustrated in figure 6.
There may be some nodes receiving NS messages from
neighboring nodes,but no RAs fromleader nodes.These
so-called isolated nodes are not within the scope of any
leader nodes and,subsequently,do not have a site-local
address.The easiest way to solve this would be,to elect
all isolated nodes to leader nodes.A better solution is,
to extend the subnetwork to these nodes.The MANETFigure 6:A multi-homed node

Leader Nodes
2 2
3 3
multi−homed node
option contains the Isolated Node Flag,which tells other
nodes in the scope about an isolated node.Subsequently,
the nodes on the edge of the subnet relay RA and NDP
messages to and from the isolated node by manipulating
the hop limit field in the same way the leader node does
for the RA messages.Nodes receiving the RAs from the
leader node with a hop limit of 0 learn that they are edge
In this section,possible interaction of routing protocols
with the autoconfiguration mechanisms described above
are outlined.We specifically refer to the hierarchical
routing protocol LANMARK [17],which is a mixture of
landmark routing [18] and a “scoped” link-state routing
protocol,like,e.g.,the Fisheye State Routing (FSR) pro-
tocol [19].Landmarks are special nodes,whose presence
is announced within the entire network.They might be
compared to leader nodes used in our approach.
Nodes use a proactive link-state protocol within a lim-
ited area,the so-called scope.Each node defines its scope
as the area,which comprises all nodes with a distance less
or equal to r
hops.This definition is directly utilized in
our approach.Routes to destination nodes outside the
scope are marked as unknown.Subsequently,packets to
these destinations are sent toward the corresponding land-
mark.The closer the packet gets to the landmark and,
therewith,to the destination node,the more accurate is
the routing information.
If the LANMARK routing protocol is used,the leader
nodes should correspond to the landmark nodes.This is
desirable,because one election mechanism can be used
for building the hierarchy of both,the address autocon-
figuration and the routing protocol.Furthermore,routing
should benefit from the hierarchical address structure by
using the subnet ID of the site-local address for aggre-
gation.In LANMARK,the corresponding landmark is
identified by a subnet IDof the destination node.Further-
more,the DADof subnet IDs could be combined with the
need of some hierarchical routing protocols to announce
the presence of hierarchical components,e.g.,landmark
nodes.However,this would violate the requirement of
being independent fromthe routing protocol.
Further optimizations are possible,if the NS messages
exchanged during DAD are used to build,e.g.,a scoped
distance-vector routing table.Additional routing control
information could be encapsulated in NDP as an option.
The hierarchical address structure established by our
system may be applied to non-hierarchical routing pro-
tocols as well,by considering the site-local addresses as
a flat address space.Nevertheless,they are not able to
benefit fromaggregation possibilities offered.
The basic goal of this paper was the applicability of the
IPv6 Stateless Address Autoconfiguration and the Neigh-
bor Discovery Protocol to mobile ad-hoc networks.Sev-
eral drawbacks have been discovered,that prevent these
protocol mechanisms from being applied in mobile ad-
hoc networks.
Some extensions to the Neighbor Discovery Protocol
have been proposed that try to overcome these shortcom-
ings.In order to make autoconfiguration scalable to large
scale mobile ad-hoc networks,a hierarchical approach is
applied.For building and maintaining the hierarchical
structure,we proposed a simple leader election mecha-
nism.Our system can cope with the dynamic nature of
mobile ad-hoc networks and with network partitioning
and merging.It is independent of the routing protocol,
although optimizations are possible in conjunction with
special routing protocols.
[1] “The Monarch Project”,Carnegie Mellon Univer-
[2] “The Wireless Adaptive Mobility (WAM) Lab-
oratory”,University of California Los Angeles,
[3] S.Thomson and T.Narten,“IPv6 Stateless Address
Autoconfiguration”,Dec.1998,RFC 2462.
[4] T.Narten,E.Nordmark,and W.Simpson,“Neigh-
bor Discovery for IP Version 6 (IPv6)”,Dec.1998,
RFC 2461.
[5] T.Narten “Neighbor Discovery and Stateless Auto-
configuration in IPv6”,IEEE Internet Computing,
Volume 3,Issue 4,July-Aug.1999 Pages 54-62
[6] Charles E.Perkins,Ad Hoc Networking,2000,
[7] J.Broch,D.A.Maltz,D.B.Johnson,Y.-C.Hu,
J.Jetcheva,A Performance Comparison of Multi-
Hop Wireless Ad Hoc Network Routing Protocols,
In Proceedings of ACM/IEEE MOBICOM’98,Dal-
las,TX,Oct.1998,Pages 85-97
[8] S.R.Das,C.E.Perkins,E.M.Royer,Performance
Comparison of Two On-demand Routing Protocols
for Ad Hoc Networks,In Proceedings of IEEE IN-
FOCOM 2000,Tel Aviv,Israel,Mar.2000,Pages
[9] Yih-Chun Hu,Jorjeta G.Jetcheva,David B.John-
son,David A.Maltz,“The Dynamic Source Rout-
ing Protocol for Mobile Ad Hoc Networks”,Mar.
2001,Internet Draft:draft-ietf-manet-dsr-05.txt.
[10] Charles E.Perkins,Elizabeth M.Royer,and
Samir R.Das,“IP Address Autoconfiguration for
Ad Hoc Networks”,2000,Internet Draft:draft-ietf-
[11] Jung-Soo Park,Yong-Jin Kim,and Sung-Woo
Park,“Stateless Address Autoconfiguration in Mo-
bile Ad Hoc Networks Using Site-Local Address”,
2001,Internet Draft:draft-park-zeroconf-manet-
[12] A.J.McAuley,K.Manousakis “Self-Configuring
Networks”,MILCOM 2000.21st Century Military
Communications Conference Proceedings,Volume
1,2000 Pages 315-319
[13] A.Misra,S.Das,A.McAuley,S.K.Das “Autocon-
figuration,registration,and mobility management
for pervasive computing”,IEEE Personal Commu-
nications,Volume 8,Issue 4,Aug.2001 Pages 24-
[14] J.Bound,M.Carney,C.Perkins,and R.Droms,
“Dynamic Host Configuration Protocol for IPv6
(DHCPv6)”,2001,Internet Draft:draft-ietf-dhc-
[15] R.Hinden and S.Deering,“IPv6 Address Architec-
ture”,July 1998,RFC 2373.
[16] Sze-Yao Ni,Yu-Chee Tseng,Yuh-Shyan Chen,and
Jan-Ping Sheu,“The Broadcast Storm Problem in
a Mobile Ad Hoc Network”,1999,Proceedings of
[17] Mario Gerla,Xiaoyan Hong,Li Ma,and Guangyu
Pei,“Landmark Routing Protocol (LANMARK) for
Large Scale Ad Hoc Networks”,May 2001,Internet
[18] P.F.Tsuchiya,“The Landmark Hierarchy:a new
hierarchy for routing in very large networks”,1998,
Computer Communication Review,vol.18,no.4,
[19] Mario Gerla,Guangyu Pei,Xiaoyan Hong,and
Tsu-Wei Chen,“Fisheye State Routing Protocol
(FSR) for Ad Hoc Networks”,Nov.2000,Internet