Information Technology Reference
In-Depth Information
P3:
Fidelity though redundancy - due to their physical constraints, individual nodes
are prone to failure through deliberate attack or normal malfunction. The redundancy
of nodes is used to compensate for this.
P4:
Flexibility - aimed at operating under diverse conditions with minimal structured
support, for example deployment in remote areas.
P5:
Dynamic network topology - the topology may change often.
P6:
Frequently data centric - IP addresses are not used, all nodes perform data-centric
routing.
P7:
Self-organising - network connectivity is often ad-hoc and dynamically
maintained.
P8:
Distributed computation - each node carries out simple data processing locally
and sends out the partially processed data to other nodes. The chain of partial
processing by individual nodes provides an aggregated solution.
Together, these properties have provided the catalyst for a wide range of new
applications, including environmental monitoring, disaster relief operations, military
control/surveillance and health monitoring [2].
3.2 Directed Diffusion
In addition to the distributed and dynamic nature of sensor network hardware, one
popular routing method is equally suggestive of natural immune metaphors: the
Directed Diffusion
protocol. This is a routing algorithm used to gather data sensed by
a large number of sensor nodes and disseminate to a node that requests such data [9].
Directed Diffusion works in two phases, an initial exploratory phase that is followed
by a reinforcement phase. Together these phases make up the three different stages
discussed in Fig. 1.
The requesting node, referred to as the 'sink node' may request data from one or
multiple other sensor nodes. As shown in Fig. 1(a), the sink periodically broadcasts its
'interest' packets (containing a description of the sensing task e.g. the regular reading
of a patient's blood pressures) to its neighbours. Interest packets are then propagated
throughout the whole network, resulting in creation of gradient fields representing the
possible data flow paths from the source, back to the sink as shown in Fig. 1(b). Once
the sink receives its requested data, it is then in a position to choose between its
various neighbours by reinforcing the paths deemed most advantageous, for example
based on the quality of service on the path that led to the neighbour, as shown in Fig.
1(c). As a result, though during the exploratory data packets are forwarded toward the
sink node along multiple paths, the gradient refinement process chooses the most
preferred path.
Reinforcements in Directed Diffusion come in two forms: positive and negative.
Positive reinforcement encourages data flow along a given path, and the result is that
data flows at a higher rate through the given path. In contrast, negative reinforcement
discourages data flow along given paths, thereby reducing the rate at which data is
sent through the path. The result is that the algorithms is dynamically able to tune its
performance (with respect to the data flow path) based on arbitrary criteria.
