Database Reference
In-Depth Information
where
S
o
is the subset of sensors that will be chosen for sampling,
C
(
S
o
)
and
B
(
S
o
)=
1
j
:
s
j
∈
S
o
B
j
(
S
) are respectively the total cost (or energy
required) and average confidence for sampling sensors
S
o
.Sin ethe
problem in Eq. (2.2) is NP-hard, BBQ proposes a greedy solution to
solve this problem. Details of this greedy algorithm can be found in [17].
By executing the proposed greedy algorithm, BBQ selects the sensors
for sampling, then it updates the Gaussian distribution, and returns the
mean values
v
11
, v
12
,...,v
1
m
. These mean values represent the inferred
values of the sensors at time
t
1
. This operation when performed ten
times at an interval of one second generates the result of the sensor data
acquisition query (Query 2.1).
|
S
o
|
2.4 Push-Based Data Acquisition
Both, TinyDB and BBQ, are pull-based in nature: in these systems
the central server/base station decides when to acquire sensor values
from the sensors. On the other hand, in push-based approaches, the
sensors autonomously decide when to communicate sensor values to the
base station (refer
Figure 2.3
). Here, the base station and the sensors
agree on an expected behavior of the sensor values, which is expressed as
a model. If the sensor values deviate from their expected behavior, then
the sensors communicate only the deviated values to the base station.
2.4.1 PRESTO.
The Predictive Storage (PRESTO) [41] sys-
tem is an example of the push-based data acquisition approach. One of
the main arguments that PRESTO makes against pull-based approaches
is that due to the pull strategy, such approaches will be unable to ob-
serve any unusual or interesting patterns between any two pull requests.
Moreover, increasing the pull frequency for better detection of such pat-
terns, increases the overall energy consumption of the system.
The PRESTO system contains two main components: PRESTO prox-
ies and PRESTO sensors. As compared to the PRESTO sensors, the
PRESTO proxies have higher computational capability and storage re-
sources. The task of the proxies is to gather data from the PRESTO
sensors and to answer queries posed by the user. The PRESTO sensors
are assumed to be battery-powered and remotely located. Their task is
to sense the data and transmit it to PRESTO proxies, while archiving
some of it locally on flash memory.
Now, let us discuss how PRESTO processes the sensor data acqui-
sition query (Query 2.1). For answering such a query, the PRESTO
proxies always maintain a time-series prediction model. Specifically,
PRESTO maintains a seasonal ARIMA (SARIMA) model [60] of the
