Image Processing Reference
In-Depth Information
As is shown in Figure ., SPIN works in three stages. When a node has new sensor data, it sends
an ADV message to its neighbors with the relative metadata (ADV stage). When neighbors receive
the ADV message, they check whether the advertised data has already been received (or requested). If
not, the node sends a REQ message to the source (REQ stage). Finally, the source of the ADV message
responds to the REQs with a DATA message, containing the advertised data. Upon receiving data,
a node can apply data aggregation techniques and then advertise the aggregated data. Furthermore,
alow-energythresholdcanbeusedtoreduceparticipationintheprotocol,i.e.,anodewillparticipate
in a stage only if its remaining energy is sufficient to complete the following stage.
As compared to classical Flooding and Gossiping protocols, SPIN features shorter dissemination
times and higher reliability. As the power consumption of the nodes during the transmission of
DATA packets is usually much higher than that needed for ADV/REQ packets, SPIN also provides
lower energy consumption than flooding. However, the advantage in terms of energy consumption
becomes less pronounced when the amount of data is small. In fact, the energy reduction technique
adopted here decreases the amount of data to be transmitted. his is effective when data transmission
requiresmorepowerthanreceptionandidlestates.However,asdiscussedin[Abi],thisassump-
tion is not always true for sensor nodes. For this reason, a number of cross-layer routing approaches
exist, such as LEACH [Hei], which directly manage the low-power/sleep states of the nodes, thus
decreasing their duty cycles.
7.3.2 Directed Diffusion Protocol
Directed Diffusion, presented in [Int], represents one of the most important data dissemination
paradigms for WSNs, as it introduces a naming scheme, in which data generated by sensor nodes is
named by attribute-value pairs. In order to save energy, short-range hop-by-hop communication
is preferred over long-range communication to the destination, and data aggregation is performed
locally to reduce the data size before transmission.
The basic idea of Directed Diffusion is that nodes request data by sending interests. An interest
specifies a list of attribute-value pairs that describe the sensing task, i.e., type, interval, duration,
sensing area, etc. Data matching the interest is drawn toward the node itself.
Both sensing tasks and data sent in response to interests are named using a similar naming scheme,
based on attribute-value pair description, i.e., name and type of objects, data rate of events, duration,
geographical area, etc.
For each sensing task, the sink node (i.e., the node that originated the query) periodically broad-
casts an interest message to its neighbors. Each node has an interest cache, where several parameters
such as a time stamp, a gradient, the data rate, duration, etc. are maintained for each distinct interest.
Each gradient is a reply link to the neighbor from which the interest was received and contains a
data rate field (derived from the interval attribute of the interest) and a duration field (derived from
the time stamp and expiresAt attributes of the interest) indicating the approximate lifetime of the
interest.
Such parameters are updated every time a node receives an interest. Interest entries do not contain
information about the sink node, as data is delivered through the gradients, which specify the
addresses of the neighbors from which the interest has been received. As the nodes maintain only
local information and no topology information is needed, the scalability of Directed Diffusion is
high. In order to spread interests throughout the network, after an interest has been received a node
may decide to resend it to all or to some of its neighbors. Each node also has a data cache used to
direct interests (thus avoiding flooding) and to prevent transmission loops.
When a node receives a data message from a neighbor node, it first attempts to find a matching
interest in the interest cache. If a matching interest is found, the message can be forwarded to each
node for which it has a gradient, otherwise it is silently dropped.
Search WWH ::




Custom Search