Database Reference
In-Depth Information
(a) Without
DBP
.
(b) With
DBP
.
Figure 7.8.
Tunnel: total link-level transmissions for a sleep interval of 1500 ms [79].
(Note the difference in the
y
-axis scale.)
very close to the cost of CTP tree maintenance, regardless of the sleep
interval. A finer-grained view is provided by
Figure 7.8
, which shows
the different components of trac in the network. Without DBP,the
dominate component is message transmission and forwarding; signifi-
cant retransmissions are present for some nodes, while the component
ascribed to CTP (i.e., the beacons probing for link quality) is negligible.
When DBP is active, the number of CTP beacons remains basically un-
changed. However, because the application-level trac is dramatically
reduced, CTP beacons become the dominant component of the network
trac.
These last observations suggest that further reductions in data trac
would have little practical impact on the system lifetime, as routing costs
are dominated by topology maintenance rather than data forwarding.
Therefore, improvements are more likely to come from radical changes
at the routing and MAC layers: new, data-aware protocols need to be
designed, which will take into account the trac patterns with extremely
low data rates that emerge when data-driven data acquisition techniques
are employed.
4.2 Uncertain Data Processing
Given the fact that sensors produce values with an inherent uncer-
tainty, and that we are increasingly relying on applications that are
driven by sensor data, it becomes evident that ecient and effective
processing of uncertain WSN data series is a relevant research direc-
tion.
Turning our attention to the three techniques we presented for un-
certain data series similarity matching (see Section 3.4), we observe
that an important factor for choosing among them is the information
