01-08-2013, 12:58 PM
The Effects of Mobility on Multicast Routing in Mobile Ad Hoc Networks
Effects of Mobility.pdf (Size: 196.58 KB / Downloads: 15)
Abstract
Performance evaluation of ad hoc routing proto-
cols typically depends on simulation, since the deployment of ad
hoc networks is still relatively rare. However, past evaluations
of multicast routing protocols have utilized a single, simple
mobility model, and thus do not capture the variety of mobility
patterns likely to be exhibited by ad hoc applications. In this
paper, we explore the impact of several different mobility
models on multicast routing performance, using techniques that
have been demonstrated to be effective for unicast routing
protocols. We demonstrate that three key mobility metrics
help to explain the performance variations that we observe for
flooding, ODMRP, and ADMR. In addition, we study a high
density, high traffic scenario and find that ODMRP copes with
this extreme situation much better than ADMR.
INTRODUCTION
Mobile ad hoc networks have numerous practical appli-
cations, such as emergency and relief operations, military
exercises and combat situations, and conference or classroom
meetings. Each of these applications can potentially involve
different mobility patterns, with movement dependent on
interactions among participants and the environment. For
example, in a search-and-rescue operation, individuals may
fan out to search a wide area, each moving independently
in a confined area. In a battlefield, however, the movement
of soldiers is heavily influenced by the movements of their
commander. Similarly, the environment can influence move-
ment, such as cars moving on a freeway or patrons in an
exhibit hall moving among a selected group of displays.
RELATED W ORK
Simulation based-evaluation of ad hoc routing protocols
depends on mobility models that characterize the movement
of mobile users [3]. Routing protocols typically establish
paths over which packets will be sent. Mobility breaks those
paths and hence disrupts the communication, exerting a load
on routing protocols. The location, duration and frequency of
these disruptions varies with the pattern of user movement.
In many mobility models, each node moves independently
from the others. Often, the speed and direction are chosen
randomly, as with Brownian Motion [9], Random Gauss-
Markov [18], Random Waypoint [15], and Random Direction
[21]. Several models restrict the direction in which nodes
may move. For example, Hu and Johnson use a Column
model, based on a design suggested by Sanchez [23], in
which nodes move with randomly-selected velocities within
a column formation [9]. Tian et al. explore graph-based
mobility, designed to model the constraints of real-world
locations, such as trains connecting cities [24]. In this model,
vertices in a graph represent possible destinations and edges
represent paths on which nodes can travel. Likewise, Davies
uses a City Section Mobility model, where several parts of a
city are modeled as streets with their speed limits [4]. When
a node has to move from one point to another, it selects the
shortest path possible with the given street constraints.
MOBILITY AND C ONNECTIVITY METRICS
Characterizing mobility models in terms of mobility and
connectivity metrics can help to explain the impact of these
models on routing performance. We implemented each of
the metrics defined in the IMPORTANT framework [1] to
determine whether they help to differentiate between the
models we are using. Of these metrics, we found that spatial
dependence and the average number of link changes are able
to differentiate between our mobility models and help to
explain multicast routing performance.
Flooding
As expected, flooding attains very high throughput at
the expense of high transmission overhead (Figure 5 and
Figure 6). This is due to each node forwarding every non-
duplicate packet it receives. Note that the transmission over-
head of 7 is a worst-case scenario because there are 49 nodes
forwarding every packet and only 7 receivers.
There are two interesting cases with respect to throughput.
First, throughput is sometimes lower when the nodes are
static because the initial placement may result in partitioning.
These partitions are never healed due to lack of any mobility.
Second, the Battlefield model results in a relatively low
throughput for flooding. Because all packets are flooded, this
must be due to partitioning. This verifies the impact of the
reachability parameter discussed in Section IV and sets an
upper bound on the performance any other multicast routing
protocol can achieve.
CONCLUSIONS
We find that mobility patterns can significantly affect the
performance of a multicast routing protocol. Our results show
that the number of link changes imposed by a particular
application is a good predictor of throughput, with greater
amounts of link changes indicating worse performance. Even
when the number of link changes is small, low node density
and low spatial dependence can also degrade throughput. We
show that high values of node density, which is typically
exhibited by group-based mobility patterns, can lead to lower
delay and lower transmission overhead. However, group-
based mobility patterns can also cause partitioning, leading
to significant packet loss.