<<

. 50
( 87 .)



>>

wireless networking turnkey products, nontechnical network edge users are increasingly
becoming wireless network owners and operators. And despite the fact that technical co-
existence problems certainly arise from highly uncoordinated deployments of unlicensed
wireless systems, the technology is here to stay.
In the not so distant past, wireless networking technologies relied exclusively on signif-
icant infrastructure planning and a strict hierarchy of control for typical operations. With
the expected tremendous growth rate of wireless applications, especially in the hands of in-
dependent end users, businesses, and communities, it will be beneficial to have more adap-
tive, self-organizing technologies that robustly operate and adapt to changes (minor or se-
vere) within a network region. Mobile ad hoc networking (MANET) is one area of evolving

255
Mobile Ad Hoc Networking. Edited by Basagni, Conti, Giordano, and Stojmenovic.
ISBN 0-471-37313-3 © 2004 Institute of Electrical and Electronics Engineers, Inc.
256 ROUTING TECHNOLOGY FOR DYNAMIC WIRELESS NETWORKING


technology that can support the operation of more adaptive wireless networks. The overall
aim of this chapter is first provide a brief introduction to MANET technology and then pro-
vide some of the authors™ perspectives on several related issues as follows:

Where and when does MANET technology make sense to consider?
What are some present or envisioned applications for MANET?
What related work has gone on within the standards community?
What are some future issues and directions for MANET-related technology?


9.1 INTERNET LAYER ROUTING AND DYNAMIC WIRELESS NETWORKS

Computer and communication networks such as the Internet are multilayered, complex sys-
tems relying on many different protocols and associated algorithms for seamless, reliable
operation. As networks extend beyond direct-link connection cases (e.g. a local area net-
work in which all nodes are logical single-hop neighbors), there is a need for some function
to forward traffic on the behalf of source systems to destination systems that are out of di-
rect connectivity range. Within the Internet Protocol (IP) suite, IP routing technology is typ-
ically used to direct the forwarding of such traffic. Several of the technical fundamentals of
routing involve how protocols find, manage, and use multihop paths for forwarding infor-
mation on the behalf of specific end systems to particular destination systems.
The global internetwork (i.e., the Internet) routing system must adapt on some reason-
able time scale to changes and failures in the network infrastructure, and it must scale to
support many billions of end systems. Thus far, the Internet routing system has evolved to
largely meet these expectations, yet the proliferation of inexpensive wireless technologies,
portable computing, and the information-hungry nature of our increasingly mobile soci-
ety pose new challenges and opportunities. New design challenges are evident at the
edges of the Internet infrastructure, where wireless technologies are being rapidly de-
ployed in the hands of users, organizations, and communities to expand local Internet con-
nections, and provide network service for a wide variety of information devices and users.
The Internet Protocol (IP) core design tenets”connectionless networking and packet-
based forwarding”are ideally suited for use in highly dynamic contexts such as mobile
wireless. Yet, new technology developments and design extensions that better address and
meet the unique challenges and opportunities of wireless operation are needed.
MANET can enable improved dynamic wireless operation by addressing routing tech-
nology improvements within this context. We define MANET operational regions as col-
lections of wireless network platforms or “nodes,” where nodes may organize and main-
tain a routing infrastructure among themselves in a relatively arbitrary fashion. Due to the
fundamental dynamic nature of wireless network communications, it is not necessary that
nodes be in motion for this to be a valuable capability, but the general design assumption
is that relative node mobility should be directly supportable. MANET nodes are enabled
with the potential for wireless-compatible routing capabilities. With these technology en-
hancements, nodes can more effectively monitor and adapt to changes in the local neigh-
borhoods and across MANET topological regions of operation. Figure 9.1 shows a simple
example of the need for dynamic routing when a wireless topology change occurs. In this
case, routing provides the functionality to forward traffic from node A to node F. As dy-
namics cause the achievable network topology to change (e.g., node movements, wireless
257
9.1 INTERNET LAYER ROUTING AND DYNAMIC WIRELESS NETWORKS




Figure 9.1. Dynamic routing in a changing topology.


link failures), valid routes must be discovered and maintained in order to forward network
data to the desired destination, node F in this example. This capability is no different from
the general goal of IP layer routing, but the underlying design assumption of wireless in-
terfaces and possibly mobile routing nodes presents increased technical challenges.
Overall, MANET deployments have been envisioned in many different scenarios and
on many different scales. In considering use of a particular MANET protocol or approach,
it is important to be cognizant of operational parameters that can directly affect suitability
and performance. Some operational parameters that affect overall performance and scala-
bility include number of peer routing nodes, type and degree of link dynamics, expected
user traffic patterns, network density, and lower-layer technology characteristics. The in-
terplay of all these parameters and their relative performance effects can be quite complex
and different for each routing approach under consideration. Aware of the oversimplifica-
tion and somewhat arbitrary nature of the following terminology, we define some rough
scalability regimes based on the number of peer routers within a region to aid discussion:

Small-scale (i.e., 2“29 nodes)
Moderate-scale (i.e., 30“100 nodes)
Large-scale (i.e., 100+ nodes)
Very large scale (i.e., 1000+ nodes)

We are applying this terminology within a single operating region, and, therefore, a de-
ployment of 10 moderate-scale MANET network operating regions consisting of 100
nodes each (1000 total nodes) is not a very large scale MANET in this terminology, but is
a considered collection of moderate-scale networks. At the time of this writing, MANET
operational experience has been gathered in small- to moderate-scale routing region
cases. Early experimentation and use of various routing schemes has been performed on a
variety of working hardware and software operating systems. In addition to working
MANET systems, a large number of independent simulation models have been developed
and numerous performance studies have been performed, mostly at the moderate-scale
and some at the large-scale level across a wide variety of protocol types and within a
range of available network simulation packages [9“11]. In the authors™ opinion, there re-
mains a growing amount of promising future technical work to be done in the area of cre-
ating large to very large MANET region technology, but even with some early implemen-
258 ROUTING TECHNOLOGY FOR DYNAMIC WIRELESS NETWORKING


tations much of this scalability work remains at a research stage with many practical is-
sues regarding performance to be further explored. On the other hand, small-to-moderate-
scale mobile network applications are reaching a level of understanding and maturity to
be operationally viable in a wide variety of scenarios.
Before we address further performance and application issues related to MANETs, let
us discuss some of the motivation behind developing MANET technology specifically for
the Internet protocol suite.

9.1.1 Why the Internet Protocol (IP) Layer?
The following fundamental question has been asked many times and deserves considera-
tion here before we continue: “Why is it especially important to solve mobile routing
problems at the IP layer?” It is clear that mobility-enhanced routing functionality can be
developed at lower layers of a protocol stack (i.e., below the IP layer). Examples of such
prior work include HIPERLAN 1 and lower-layer cluster-based routing [12, 13]. Such
subnet convergence technology can provide a logical local area network (LAN) appear-
ance to the IP layer and handle a degree of packet relaying. To the authors, it becomes not
a simple question of where the technology belongs. Lower-layer approaches remain valid
engineering design options, and improving routing functionality at different layers of a
protocol stack can support different architectural needs. There remain several key reasons
to consider and promote fundamental dynamic routing enhancements at the IP layer.
First, the main reason for doing work at the IP layer is to better support heterogeneity
and networked interoperability of lower-layer technologies”to continue to build a net-
work of networks. Multiple link layers may be simultaneously deployed, the physical com-
position of which can form a unified, logical whole via routing at the IP layer, as illustrat-
ed in Figure 9.2. The number of commercially available wireless technologies and devices




B D F
A G
C E
Heterogeneous Topology
at the IP Layer
logical
physical
Physical Topology of
Link Type X (e.g., 802.11)



MANET node F
D
B MANET node
w/ 2 interfaces E
C w/ 1 interface
G
A




Physical Topology of Link Type Y (e.g., IR)


Figure 9.2. A heterogeneous mix of MANET node and interface types.
259
9.1 INTERNET LAYER ROUTING AND DYNAMIC WIRELESS NETWORKS


continues to increase over time [e.g., infrared, 802.11, Bluetooth, ultrawideband (UWB)].
We expect lower-layer wireless technologies to continue to evolve and vary in use and
popularity, yet IP tends to remain as the fundamental internetworking glue working across
multiple technologies. As this understanding permeates the industry, we predict that link
layers will be more expressly designed to plug into the underbelly of an IP network. Sec-
ond, any IP layer software development often capitalizes on the rich variety of existing
networking support already available within IP protocol software stacks and operating
systems, thereby reducing development and deployment costs, and simplifying redesign
and replication efforts. Software modifications and upgrades are also easily performed.
Third, many wireless deployment applications require an IP routing approach at some lev-
el of the architecture, so improvements relating to wireless performance, mobility, and ro-
bustness are generally desirable.

9.1.2 What Do MANET Nodes and Networks Look Like?
MANET-enabled devices can come in a large variety of forms, but there is basic function-
ality that all MANET nodes at least partially possess. Figure 9.3 shows two types of
MANET nodes: first, a simple MANET node is shown as a locally contained computer
host with a single wireless network interface; second, a more connected MANET node is
shown supporting multiple wireless or wired interfaces and potentially connected network
segments. This second node is also potentially providing direct routing support for a set of
attached hosts or network prefixes as well.
The overall functional description of a MANET wireless routing node is largely con-
sistent with the conventional view of Internet routing devices (see Figure 9.3B), except for
some interface behavioral design assumptions. One such difference is that a single inter-
face device is acceptable as a common form of a MANET routing node (See Figure
9.3A). This is true because in a broadcast-oriented wireless interface, a node A will have


Attached Wireless Interface
Hosts/Networks


1
Wireless Interface
MANET
3 NODE
1
2
MANET
NODE


Second Wireless Interface
A: SIMPLE MANET NODE

B: MULTIPLE INTERFACE MANET NODE
• Single wireless IP interface
• No attached local networks
• Multiple IP interfaces (wireless or wired)
• MANET routing supports attached
local networks and hosts

Figure 9.3. Types of MANET node configurations.
260 ROUTING TECHNOLOGY FOR DYNAMIC WIRELESS NETWORKING


neighbors on the common physical interface that another node B may not have, due to rel-
ative node positions or other interference and propagation effects. Thus, unlike many
wired protocol designs, individual nodes may and often should forward downstream traf-
fic on the same interface on which a data or control packet was received. This forwarding
is required on behalf of other nodes out of range of a sending upstream node. There are
other scenario-dependent factors that make MANET network interfaces potentially differ-
ent from conventional wired network interfaces as follows:

Relative neighbor motion, environmental, and distance effects
Dynamic local noise and interference (possibly self-induced)
Time-varying communication channels due to intentional or unintentional causes
Asymmetric neighbor links often exist
Lower-layer wireless protocol behaviors (retransmit buffers, reliability, etc.)

The main point here is that MANET neighbor interface links can and often do behave and
appear quite different than those assumed in the wired networking world.
In addition to enhanced wireless mobile routing, connectivity and interoperability with
the rest of the Internet infrastructure is an important area of consideration. At present,
working MANET experimental networks often operate as Internet stub routing areas.
We now consider how a set of deployed MANET nodes might appear within a larger
network context. Once again, there exist a large variety of possible application and de-
ployment scenarios, and we discuss only a few general examples. In its present stage of
engineering experience, MANET technology is perhaps most commonly suitable as a
means to provide Internet connection for a wireless routing region at the edge of the Inter-
net”to extend network range or to provide robust adaptation to infrastructure dynamics.
Figure 9.4 illustrates a MANET stub area connected through a single Internet access
router. In this case, the MANET routing is functioning solely at the edge of an existing In-
ternet infrastructure to support the seamless extension of that infrastructure into a more
dynamic, wireless environment. Existing, low-complexity MANET routing solutions that
have recently emerged from research into development can be used effectively within




Figure 9.4. Basic MANET Internet extension.
261
9.1 INTERNET LAYER ROUTING AND DYNAMIC WIRELESS NETWORKS


such a stub network scenario. In addition, multiple Internet points of attachment can often
be supported but are not shown in Figure 9.4. Since, in this scenario, connection and ac-
cess to the greater Internet is assumed to be of primary importance, addresses within the
area can be managed and even autoconfigured through protocol and management tech-
niques focused at the edge of the MANET area.
Stub area operation simplifies routing interoperability issues and forwarding policies

<<

. 50
( 87 .)



>>