<<

. 51
( 87 .)



>>

and is often the most typical application scenario of present MANET experimentation and
deployment. Recent proliferation in working MANET-based experiments and demonstra-
tions has been made possible by the variety of working prototype implementations avail-
able for general use [8]. The use of a globally advertisable prefix or set of prefixes within
the MANET stub is quite straightforward and simplifies the router border gateway adver-
tisement and routing table exchange issues. This stub scenario also allows simpler ap-
proaches to MANET autoconfiguration, and at the Naval Research Laboratory (NRL) we
have demonstrated a number of working small-scale networks demonstrating MANET au-
toconfiguration, as depicted in Figure 9.5.
In the working example, node A serves as a MANET gateway node and provides de-
fault routing to the larger Internet. Node X is a wireless node that has been previously au-
toconfigured and is now operating as a functional MANET router and simultaneously as a
Dynamic Host Configuration Protocol (DHCP) relay agent for node A. A joining node Y
now enters the network and obtains a MANET region routable address via associated
neighbor DHCP relay agent functionality. Once it obtains a valid routable address, node Y
can begin participating in a MANET routing protocol and also begin hosting related ser-
vices such as a DHCP relay function. At NRL, we are also demonstrating related autocon-
figuration operations for IPv6-oriented wireless networks and are looking at the integra-
tion of stateless address autoconfiguration and distributed service discovery and
interaction methods involving anycasting. Our work and related work of others is present-
ly ongoing but will likely result in multiple approaches and techniques to autoconfigure
MANET networks in the future.
As demonstrated in Figure 9.5, a small-scale MANET can be completely autoconfig-
ured including the establishment of addressing, service discovery, and dynamic MANET



Stateful
Server
A (e.g., DHCPv4/v6)
Manet GW




Stateless or Stateful
Joining
Address Configuration
Node
Wireless
Ad Hoc
Network Relay
X Y
Agent


Figure 9.5. MANET stub autoconfiguration example.
262 ROUTING TECHNOLOGY FOR DYNAMIC WIRELESS NETWORKING


routing. By using simple conventional-type approaches for autoconfiguration combined
with a MANET routing protocol, an adaptive wireless network can be realized for many
practical applications (e.g., a routable home network). More complex MANET address
management and configuration schemes can be devised, and may be of interest, but are
not discussed further here since these concepts are evolving and they may not be neces-
sary for many practical applications.
Beyond the more common Internet extension application for MANETs, dynamic
MANET routing areas may also be formed independent of any external network connection
to support completely autonomous networked operation. Examples include the use of such
techniques within an area to perform ad hoc collaboration and computing in an emergency
or disaster relief scenario. Other such uses could be in temporary, dynamic collaborative
business, private intranet, or robotic applications. In the case when an infrastructure con-
nection is lacking or not of primary application interest, the autoconfiguration and ad-
dressing issues become more challenging. In this case, more peer-to-peer type approaches
may be useful and are still being discussed and evolved within the technical community.
Only a few years ago, few actual working implementations of MANET protocols exist-
ed in practice. Yet, at the time of this writing, a wide variety of MANET prototype soft-
ware has been demonstrated on a variety of diverse end platforms. A key point relating to
this is that interoperating MANET nodes can always be a heterogeneous collection of de-
vices and platforms of different capabilities and uses. Figure 9.6 shows a number of dif-
ferent sublaptop to handheld devices presently demonstrating functional MANET routing
capabilities within a dynamic wireless topology.

9.1.3 Wireless Characteristics and Applicability
The goal of MANET routing is to provide enhanced IP routing for wireless networks, es-
pecially those that are possibly mobile or highly dynamic. The lessons learned from such
designs may also be useful in wired protocols, but the unique challenge of wireless opera-
tion provides the primary design motivation. We previously mentioned a few unique be-
havioral aspects of wireless MANET interface types. In addition, there are numerous op-
erational factors that significantly distinguish mobile wireless networks from fixed
networks including [3]:

Nominally lower capacity is typically available as compared to wired network coun-
terparts. This is becoming less of a concern in more recent applications using short-




Embedded Devices
Sublaptops PDAs (e.g., Sensor Node,
Appliance, Vehicle)

Figure 9.6. Some Example MANET prototype platforms.
263
9.1 INTERNET LAYER ROUTING AND DYNAMIC WIRELESS NETWORKS


range, high-capacity wireless communications, but there remain scenario-depen-
dent issues relating to power, spectrum, and antenna design.
Limited broadcast nature of some wireless multiple access media. Again, this re-
lates to the use of a single interface for relaying and connecting devices out of range
but on the same physical interface. Many existing wired routing protocol designs
assume that this type of forwarding should not occur.
Increased likelihood of channel interference and congestion detection problems.
This may be due to bandwidth constraints, hidden-terminal problems, frequency re-
strictions, or channel access techniques.
More frequent topological changes. This may often be due to node mobility, chan-
nel propagation effects, resource failures, power control, or antenna dynamics.
Higher loss rates (e.g., due to interference, fading, congestion or network dynamics)
Potentially higher delays and jitter (e.g., due to lower transmission rates, link layer
retransmissions, use of long propagation delay links, or dynamics)
Lower physical security of media (e.g., due to lack of physical control over media)

There is a significant history of packet radio network research and development going
back to the early 1970s [14]. However, in the past, mobile wireless network designs were
often looked upon as homogeneous radio frequency (RF) media problems. With time and
the proliferation of numerous proprietary radio networks, the need for heterogeneous in-
teroperability across networks is becoming a pressing concern. Reflecting upon past In-
ternet technology development, it is clear that support for a heterogeneous mix of tech-
nologies and devices is one of the great successes of IP. In the near future, computing and
network routing devices may typically have multiple wireless media interfaces (e.g., ultra-
wideband, Bluetooth, Zigbee, 802.11 variants, cellular). This proliferation of ubiquitous
wireless devices is expected to continue to evolve with many newer technologies to
choose from over time. IP routing-layer technology provides multihop relaying and dy-
namic internetwork connection support. In this broad sense, IP technology has supported
and will continue to support both wired and wireless infrastructures.
As we face increasingly embedded and widespread wireless network technology, wire-
less routing performance that deals with increasing temporal and topological dynamics is
a key enabler. We wish to emphasize that increased dynamics may not always result from
mobility, and, therefore, mobile systems are not the only context suitable for applying
MANET technology. Dynamic link conditions due to other system effects are often quite
evident in wireless networks, even when the nodes are static or quasistatic. In fact,
MANET approaches will likely work equally well or better than existing standard routing
when used in quasistatic wireless applications involving routing meshes. Dynamics, with-
out significant motion, may be expected in deployed cooperatives or community network
grids, where nodes may come or go at random, or in networks where energy conservation
or power cycling may be an issue. Regardless of the operational reason, the ability to man-
age and adapt to expected change with a high degree of robustness and efficiency is as-
sumed to be a fundamental desired property.


9.1.4 Networking with Small Devices
MANET technology, because it provides dynamic and mobile wireless network support,
is naturally being considered for use in embedded devices, including human wearable and
264 ROUTING TECHNOLOGY FOR DYNAMIC WIRELESS NETWORKING


portable devices. As illustrated in Figure 9.6, devices such as compact laptop computers,
personal digital assistants (PDAs), and embedded computing systems have been demon-
strated running varieties of MANET routing technology. This technology achievement en-
ables the ad hoc formation and maintenance of dynamic wireless infrastructures even
among small, embedded devices. However, there are a number of issues related to the use
of small, portable devices that deserve consideration when developing and selecting ap-
propriate MANET technology or modes of operation including:

More limited energy (e.g., possibly battery-/solar-operated devices)
More limited computing power
More limited memory
Increased interference and dynamics (nearfield RF effects)

Limited energy can be a critical operational consideration. For instance, if a battery-
powered device is naturally power cycling (e.g., into and out of a dormant mode) this may
need to be considered as an effect in the design assumptions of a routing protocol. Ques-
tions of importance may include the following:

What are the related energy costs of transmit, receive, and dormant modes of a node?
Is transmit power control desirable to adjust the number of active neighbors in a
topological region?
Is it desirable to conserve the energy of the network as a whole, the individual de-
vice, or both?

Besides the potential energy conservation issue with embedded devices, there is also an
issue of complexity. A heavyweight, complex routing protocol that may work well on a
modern laptop or desktop computer with significant memory and processing capability is
not necessarily the best approach for an embedded network processor or PDA application.
The fact that embedded computers vary in their memory and computing capabilities”
some rivaling desktop systems of several years ago”lead us to several conclusions on
this issue. First, there is likely more processing and memory freedom in algorithm and
protocol design than there was in the design space of the early Internet days, even for
many embedded computing applications. Yet, second, while Moore™s Law seems to equal-
ly apply to embedded computing capability growth (albeit as a lagged variant), we still
need to be cautious as there is an increasing cost to be paid for burdening these precious
local resources with overly burdensome protocol and software designs, especially since
battery capacity improvements are not following Moore™s Law.


9.2 MANET TECHNOLOGY APPLICATION

This section discusses more select scenarios that the authors are aware of being demon-
strated successfully using MANET technology and future applications that are being con-
sidered for adaptation and use. This list of uses and applications is not intended to be com-
plete and there are other applications for MANET technology. Likely, there are many that
are not presently realized or envisioned by the authors, not unlike the case of the early In-
ternet. To begin, we reemphasize that MANET is not solely intended for disconnected au-
tonomous operation or scaled scenarios (e.g., hundreds or even thousands of cooperating
265
9.2 MANET TECHNOLOGY APPLICATION


wireless nodes in a region). These are interesting and important potential application
areas, but we should not ignore the important and more practical basic infrastructure en-
hancement applications of MANET. As an example, the general use of MANET technolo-
gy as a basic wireless stub network extension, as illustrated in Figure 9.4, has many poten-
tial common applications and for small-to-moderate-size network scenarios is a highly
practical approach. Beyond this basic application, we now discuss a set of other potential
application areas and related issues.

9.2.1 Hybrid Infrastructure Extension
In the authors™ opinion, a potentially widespread use of MANET solutions is likely to be
in supporting low-complexity, effective hybrid infrastructure extensions where needed or
desired. Many MANET solutions are low complexity and are beginning to become avail-
able in a wide variety of implementations. A simple example of an application that the au-
thors have actually deployed on a small scale is a dynamic enhancement to a home or
campus wireless networking environment. A typical campus deployment process today in-
volves wireless area surveys to decide how many access points to deploy and where to de-
ploy them. This can work reasonably well if network wireless access devices are deployed,
managed, and configured properly to cover all communication areas of interest and all
possible scenarios of desired network communications. Some node handoff techniques
are designed for distributed access points, but more direct dynamic IP routing and IP
router device association and flexibility can offer advantages over proprietary and more
limiting techniques within an operational region. Typically, even within a well-architected
fixed-backbone system there are problem coverage spots, dynamic outages, as well as
short term and long term-dynamics that should be addressed with more flexibility. The
use of MANET technology to provide extended service allows low-cost, low-complexity
dynamic adjustments to provide coverage regions and range extensions away from the
more fixed infrastructure backbone networks.
More recent MANET protocol developments support options to allow certain MANET
routing nodes to be preferred over others in a neighborhood [15, 16]. In the extreme, a
node can participate in dynamic neighborhood discovery mechanisms of a MANET pro-
tocol but may be excluded through low preference from becoming a forwarding or routing
node. Through the use and management of a such functions, a fixed wireless infrastruc-
ture (preferred and close to the Internet backbone) can be deployed in a dynamic environ-
ment with more passive MANET nodes participating in local neighbor exchanges and dis-
covery only. If management policy or preference allows, these more passive MANET
nodes can provide range extension and dynamic routing functions on an as-needed basis.
A fictitious but somewhat practical example of such a hybrid grid deployment is shown in
Figure 9.7, where the static building and utility pole nodes are managed wireless MANET
devices that have been well placed and/or loosely coordinated to form a wireless access
grid. The wireless access grid in this example is made up of preferred-access nodes close
to the fixed infrastructure. Other nodes in the picture demonstrate passive MANET nodes
that may be static, power cycling, or moving through the grid to gain Internet communica-
tions. In this way, nodes moving or operating on limited energy may be low-preference
routing nodes, thus providing more physical stability to the overall routing grid as well.
Also depicted is a node at the bottom right that cannot reach or discover any primary ac-
cess grid nodes and requires some range extension assistance. In this case, a previous pas-
sive MANET node, to the left of the one shown in the lower right of Figure 9.7, is provid-
266 ROUTING TECHNOLOGY FOR DYNAMIC WIRELESS NETWORKING




Figure 9.7. Hybrid MANET access grid application.




ing a limited routing function (if policy allows) for the lower right node that is out of
range of any preferred nodes. When an assisted passive node can directly link to a prima-
ry access router, the assisting node can quickly return to its passive role within the grid.
The additional routing functionality to support such a flexible capability is very light-
weight but can provide a powerful management function for hybrid systems.
The hybrid grid notion demonstrated by Figure 9.7 illustrates a very practical applica-
tion of MANET technology that may be appropriate within a campus, community, robot-
ic, sensor, or localized business application. The passive MANET concept easily supports
network devices that require dynamic routing support but are not preferred or allowed
routers (e.g., intermittent battery-powered PDA connections, known highly mobile or dis-
advantaged nodes, etc).
It is in contexts such as hybrid applications that MANET technology can most effec-
tively contribute toward the oft-cited vision of “Ubiquitous Computing,” a research field
originated by Mark Weiser and described in his seminal paper [1]. To quote from the pa-

<<

. 51
( 87 .)



>>