Database Reference
In-Depth Information
is the data-centric storage [19 20 21 22 23 24 25]. In data-centric storage,
the location to store a piece of data is determined by a set of attributes
of the data. The benefit of such scheme is that all the related data could
be stored together. Sophisticated algorithms are needed to determine
where a piece of data should be stored so as to balance the storage cost
of all nodes. A. Omotayo et al. [18] build a model for a data-centric
storage scheme with respect to the energy cost for storing and retrieving
data in WSNs. Suppose a piece of data
d
is generated by node
n
src
and is stored at node
n
dest
. The total cost of storing d contains three
components: (i) Reading
d
from the memory of
n
src
, (ii) transmitting
d
to
n
dest
, and (iii) writing
d
to the memory of
n
dest
. The total cost of
retrieving d contains three components: (i) Routing the retrieval request
to
n
dest
, (ii) reading
d
from the memory of
n
dest
, and (iii) returning
d
to
the base station.
Suppose a node
n
i
stores part of its data (with size
M
i,j
)atanother
node
n
j
. The energy cost of storing and retrieving data increases along
with the following parameters: i) the distance of ni from the base station,
ii) the size of
M
i,j
, iii) the distance between
n
i
and
n
j
, and iv) the
distance of
n
j
from the base station. These critical parameters are crucial
to reducing the energy cost for data-centric storage.
5. Data Acquisition and Aggregation
This section surveys data acquisition and aggregation techniques. Sec-
tions 5.1 and 5.2 introduce query models and general frameworks for
data aggregation and acquisition. Section 5.2 surveys ecient algorithms
for specific applications. Section 5.3 investigates secure aggregation in
WSNs. Section 5.4 discusses an extension of the general frameworks to
support ecient in-network joins.
5.1 Query Models
Since sensors acquire samples of the environmental parameters peri-
odically, the data from the WSNs are streams. There are two query
models in WSNs: push-based and pull-based. In the push-based model,
the user registers a continuous query at the base station
n
0
.Thequery
is then disseminated by
n
0
and stored in the network for a relatively
long period of time, during which the sensors continuously generate the
results that satisfy the query and push them to the base station. This
model is the most common and practical one in WSNs. A typical query
over the WSN contains the following information: (i) The sampling fre-
quency: how often the sensors take samples, e.g., once per minute, (ii)
the affected attributes: which attributes should be sampled, e.g., tem-
