1.1 Introduction 3
which are shifting with time. These models may be used for tracking changes in
a network environment and subsequently determine the future environment of the
network. If we are able to anticipate the changes in the network environment, we
will be capable to carry out several schemes to cope with different spots such as
frequently changing topology, more static routes, high delay routes. For illustration,
the prediction of changes in connectivity can help us to choose more stable routes. If
the prediction of the future neighbours of a client with good accuracy can be imposed
with the routing algorithm, the router will estimate the mobile nodes’ available time
on the route and choose better paths with longer route expiration time.
In this book, we take up the challenge of identifying the network parameters
dependent on time and possess a large impact on MANET working. The parameters
identified are neighbour count, link load, path length, cluster count and delay. The
number of neighbours of a node is important data for several network services such as
network connectivity, routing, congestion control, topology construction, protection.
Neighbour counts of a node are parameter changing continuously due to change in
physical locations of mobile clients. So this is a potential candidate to put research
effort on and model in order to improve network functionality.
The link load is another parameter which exhibits high dynamism with time. It is an
indication of congestion across a link. This parameter may be used for supplementing
Media Access Control (MAC) protocols, so that they can take care of congestion on
the link.
The path length between a source–destination pair which is the total number of
links between the said source–destination pair is another parameter worth of engaging
needs high attention. The mobile nodes are often configured to work on a smaller
transmission range due to limited battery power and better throughput. So, most of
time, the length of path is more than one. The path length varies constantly with a
change in the neighbourhood of either source, destination or any intermediate node
on the path. For weighted clustering algorithms, predicting weight can be very useful.
A good deal of savings can be made in weight-based clustering scheme if a node can
calculate the weights of another client. So, we consider the weights of the nodes in
weighted clustering technique as our next parameter to obtain a theoretical account
for it.
The end-to-end delay happens to be another such parameter which needs more
attention to make MANET applicable to various real-time applications such as mul-
timedia communication. Because of multihop nature and continuous movement of
nodes, end-to-end delay in MANET is higher, compared to other infra-structured
network.
We have put above parameters in the framework of time series because all the
said parameters have temporal implication. They shift with time regularly. Although,
time series has been enforced in some sphere of computer networks like modelling
of Internet traffic, delay, but it has not been taken up for modelling and predicting
the said parameters. The said parameters, when modelled using time series frame-
work, exhibit a right fit with autoregressive AR(p) model of order p. The order p is
determined for each fitted model and found lying between one and three. These fitted
models have been used for forecasting the future values of the said parameters and
found to be in full accordance with the actual values validated by statistical test.