Generally, a sensor node is equipped
with processor and memory. Hence, a sensor network can perform distributed
computation and storage over all the sensor nodes. This e?±ciently reduces
the data transmission cost and the energy consumption of the sensors, also it
improves the performance of the sensor network. Actually, centralized query
processing can only be applied in the condition that sensors have su?±cient
power supply and low sampling rate.
Distributed query processing
Considering the drawbacks of centralized query processing, researchers at
Cornell propose some distributed query processing techniques [2, 3]. In their
methods, a query determines which data should be retrieved from the sensor
network and the aggregations in the query are processed in-network. Di?®erent
queries need di?®erent data and only those data related to the queries will be
retrieved. The rest of this section compares the di?®erence between centralized
and distributed query processing through an example and illustrates the
advantage of the distributed approach.
Given the sample query ???for every 30 seconds, return the average temperature
of the sensors whose readings are higher than 30?—¦C??™, we will illustrate
how to process it by the two query processing techniques, which is demonstrated
in Figure 12. In the figure, the smaller rectangles represent the data
acquired from the sensor nodes and the ellipses denote aggregate operations
which compute an average value.
Pages:
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500