Image Processing Reference
In-Depth Information
4.4.6 Data-Centric Data/Query Dissemination
Deciding how to route or disseminate data within a communication network is typically performed
using IP-style communication where nodes are identified by their endpoints which can be directly
addressed using an address that is unique to the entire network. Such addressing schemes are used
in the Internet or office LAN. Due to their application-specific nature however, WSNs tend to use a
data-centric addressing scheme where “names” are used to create an abstraction of node network
addresses. For example, instead of requesting the reading of a node with a particular ID, a typi-
cal query usually requests for an attribute of a phenomenon instead. Data dissemination schemes
generally address three main issues:
•
Howqueriesareroutedonlytotherelevantnodesfromthenodeinjectingthequeryinto
the network (Flooding a query to all nodes is not energy-efficient)
•
How results are routed back to the querying node
•
Robustness: how the scheme copes with dynamic network topologies, i.e., node failures
andnodemobility
Wereviewanumberofdatadisseminationschemeswhichusedistributedtechniquestoensure
that queries only propagate towards sensors capable of serving the injected queries.
Directed diffusion [] is one of the pioneering data-centric data dissemination paradigms devel-
oped specifically for WSNs. It is based on a publish/subscribe API where the sink injects an interest
(e.g., interest for a particular type of named data, such as the “Detection of a bird”) into the network.
Every receiving node keeps a copy of the interest in its cache. he entry in the cache also stores a gra-
dient that indicates the identity of the neighboring node that originally sent the interest. Gradients
areformedgraduallyateverynodeastheinterestpropagatesfromonenodetoanothereventually
flooding the entire network (Figure .). When a source node is able to service the interest, it sends
data back to the sink along the path of the gradients set up initially by the interest. In the event
data starts flowing toward the sink along multiple gradient paths, the source node reinforces one or
Sink
node
Sink
node
Source
node
Source
node
(1) Propagation of interest
(2) Setting up of gradients
Sink
node
Source
node
(3) Transmission of data
FIGURE .
Setting up of gradients in directed diffusion.
