Database Reference
In-Depth Information
via a multi-hop topology. Each tuple of each stream includes a time
stamp corresponding to the time it was produced. In some applications,
the lifetime of the network is much more important than the sampling
frequency. For example, in a wild life habitat monitoring application,
scientists may not be aware of how the sampling frequency will affect
the life time of the network. However, they want the network work at
least one month. The keyword ”LIFETIME” is introduced in TinyDB
to specify at least how long the network should function:
Q5:
SELECT
nodeid,movements
FROM
sensors
LIFETIME
30 days
In this case, TinyDB performs life time estimation to adjust the sam-
pling frequency (at the same time, the frequency of sending and receiving
messages), so that the remaining energy can last until the specified life
time. According to the energy costs of accessing sensors, the selectiv-
ity of the query, the expected communication rates and the remaining
energy, a sampling frequency is computed to ensure the expected life
time.
Silberstein et al. [10] propose CONstrain CHaining (CONCH) for
data acquisition. The main idea of their approach is to provide effec-
tive spatio-temporal suppression and use a minimum spanning forest of
the network for data transmission. The CONCH method is an edge
monitoring approach that exploits spatial-temporal correlations in sen-
sor readings and effectively reduces the message transmission based on
such correlation. As data transmission dominants the energy consump-
tion, CONCH exhibits considerable energy reduction during the empiri-
cal evaluation. CONCH is based on the tree topology. The base station
directly monitors the readings of a selected portion of sensors, which
report their readings to the base station. For a remaining node
n
i
,it
reports to its parent, which computes the difference on the reading be-
tween itself and
n
i
. The parent then reports to the base station if and
only if the difference changes. The base station monitors such differences
on chains of nodes. It assumes the difference between the correspond-
ing pair of nodes does not change if no report is received. Knowing
the global topology of the network, the base station is able to recover
each node's reading from the readings of the directly monitored nodes
and the chained differences. Due to the spatio-temporal correlation, the
readings of nearby sensors always share the same trend and the differ-
ences between them do not change frequently. So a large amount of
transmissions are suppressed and CONCH has good performance.
