Self-Organising Node Address Management in Ad-hoc Networks1
Stephen Toner and Donal O’Mahony
Networks & Telecommunications Research Group (NTRG) Trinity College, Dublin 2, Ireland
Tel:+353 6082336, Fax: +353-1-6772204
{stephen.toner, donal.omahony}@cs.tcd.ie
Abstract. Whilst much effort has been put into the creation of routing algorithms to handle all
帮写计算机英语论文sorts of mobility scenarios in ad-hoc networks, other fundamental issues, such as the addressesused by nodes, haven’t been dealt with adequately. This addressing problem has recently attractedincreased attention and a few proposals have been made, though often these schemes only work inlimited scenarios. In this paper we present an autoconfiguration protocol designed to work in a
mobile ad hoc network (MANET). The scheme allows joining nodes to dynamically obtainaddresses, and has been designed to efficiently manage addressing, and to handle such scenarios as
the merging and partitioning of networks. We discuss an implementation used within an emulatedenvironment and a real self-organising ad-hoc network.
1 Introduction
An ad-hoc network is one where nodes collaborate to allow communication without the required
presence of network infrastructure. Their dynamically changing membership and topology means that
specialised routing protocols are required. The lack of manual management means that autoconfiguration
is a highly desirable goal. Whilst research in ad-hoc routing protocols has been strong in
recent years, the same intensity has not been applied to other important related areas, such as node
addressing. However these routing protocols typically rely on nodes having a unique address, and
ignore this vital issue. Often nodes are assumed to have addresses configured a priori, but this is
impractical and not easily accomplished. Instead we propose a dynamic scheme for assigning and
managing addresses within ad-hoc networks. Nodes require a unique address for packets to be
delivered to the “correct” destination, and due to the routing side effects that may arise from nodes
using duplicate addresses.
In fixed IP-based networks hosts use IPv4 or IPv6 addresses, which have a hierarchical element to their
structure. An IP address typically has 2 purposes: It identifies the node, and encodes routing
information. This second point is vitally important when nodes are mobile. Traditionally this mobility
has been handled in two ways. DHCP uses servers to auto-configure nodes with a topologically correct
address as they move, so normal IP routing can be used. Also Mobile IP allows nodes to maintain a
static identity based on the “home address”, so a node is always contactable via a static address (albeit
inefficiently).
Unfortunately the use of the same combination of DHCP and Mobile IP within an ad-hoc network may
prove impractical, and may be even impossible. Because there may be no infrastructure available,
DHCP, which relies on an address server, is unable to provide a solution to the addressing problem. In
fact it is actually unnecessary to configure a topologically correct address as, unlike in the fixed
network, the address used in an ad-hoc network has no routing purposes (for most routing protocols).
Addresses are simply used as a means of identification within the ad hoc network, and so must simply
be unique within the network. Instead the ad-hoc routing protocol handles the routing, typically using
a flat address space.
An obvious solution to this problem is to provide each node with a permanent unique identifier that
could be used to identify the node within the MANET. A hardware-based solution, where every piece
of hardware has a permanent unique identifier introduces a number of problems:
1 This material is based upon work supported, in part, by the European Office of Aerospace Research and
Development, Air Force Office of Scientific Research, Air Force Research Laboratory, under Contract No.
F61775-01-WE052
• Whilst 48-bit IEEE MAC addresses are the closest thing we have to this, and are designed
