Both of these
techniques have their advantages and disadvantages, and can be applied in
di?®erent circumstances.
Centralized query processing
Centralized query processing contains two steps. First, it periodically retrieves
data from the sensor network and stores the data at a centralized
database. Second, it processes the queries on the centralized database. These
two steps can be executed at the same time.
A centralized approach is suitable for historical query processing, but this
method has many limitations. If the period of storing data to the centralized
database server is too long or the sampling frequency of sensors is too low, the
database server cannot retrieve all the data, which is necessary for query processing,
from the sensor network. For example, if the period of storing data
to the centralized database is 30 minutes, the query ???retrieve every minute
the rainfall in BeiJing??™ cannot be processed accurately. In order to answer
the query by the centralized method, the database server needs to frequently
retrieve data from all the sensors. Since this method will rapidly exhaust the
310
Chapter 12 Data Management in Sensor Networks
energy of the sensors and produce a lot of redundant data, it cannot be applied
in practice. Actually, it is unnecessary to retrieve data from the entire sensor
network just for processing a query. In the above example, instead of retrieving
data from all the sensors, the database server only needs to increase the
sampling rate of the sensors in BeiJing.
Pages:
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499