Database Reference
In-Depth Information
On receiving the signal, a node sends its current reading to the parent.
The benefit of the hierarchical thresholds is that when a reading with
value v reaches a node having its threshold
>v
, the transition stops.
It is because the global maximum value, which equals to the threshold
maintained at the root, is at least
. Since v is less than
,
v
cannot be
the global maximum value.
5.4.2 Continuous Top-k Queries
.
Top-k monitoring is a generalized version of the extreme value
monitoring. Babcock et al. [30] address the problem of monitoring top-
k values among distributed datasets. The idea is to align local top-k
lists to global top-k list through some adjustment factors. However, the
setting of the distributed datasets is quite different from that of the
WSN applications. Wu et al. [29] propose FILA for top-k monitoring
in WSNs. The basic idea is to install a local filter [
li,ui
]ateachnode
n
i
, indicating that sensor readings of
n
i
do not affect the global top-k
ranking if they fall in this filter and it is not necessary to report them. On
the other hand, once a reading falls outside of the filter, it may affect the
global top-k ranking and
n
i
reports to the base station
n
0
.
n
0
maintains
a synchronized copy for each filter. It reevaluates the filters once top-
k values change and sends the new filters with corresponding nodes.
By suppressing unnecessary updates, FILA outperforms straightforward
approach implemented in TAG.
5.5 Join Processing
In-network JOIN operation (e.g. joining two records in the WSN that
are within a specified time window) is not eciently supported in TAG.
The following shows a scenario of join in WSNs: In a volcano moni-
toring project, after noticing that the volcanic activity of the mountain
has increased, scientists want to know whether the pressures detected
have crossed a certain threshold and is continuously increasing within
some period of time. An SQL-like query
Q
6
is submitted to the network:
Q6:
SELECT
P_1.pressure, P_1.time, P_2.pressure, P_2.time
FROM
Pressure AS P_1, Pressure AS P_2
P_1.pressure > threshold
(
)
WHERE
AND
P_2.pressure > P_1.pressure
AND
P_2.time > P_1.time
AND
P_2.time-P_1.time > h
where Pressure is the relation represents the sensor data on which self-
join is performed. Specifically, two tuples of Pressure that fall in a time
window are joined. Since a sensor does not know beforehand which nodes
it may join with, a naive way is to let very node flood its tuples all over