How to process aggregation
is one of the key issues in query processing in sensor networks. This section
introduces some primary aggregation processing techniques (aggregation
techniques in short).
Centralized and distributed aggregation [10, 12]
In a centralized aggregation, a client host first gathers readings from all the
sensors and then computes the aggregate result. Figure 13(a) illustrates this
scheme. In the distributed approach, aggregation is achieved by the collaboration
of many sensor nodes. While routing the data, sensor nodes compute
the whole or partial aggregate value of the data they transfer. Finally, those
aggregate results are sent to the client host. Figure 13(b) demonstrates the
distributed approach.
Figure 14 gives an example of processing aggregation in sensor networks.
All the sensor nodes form a tree, which is called a routing tree. In the figure,
the dash lines represent connections between the sensor nodes; the solid lines
denote the edges of the routing tree which are used to transfer data from
the leaf nodes to the upper nodes while computing the aggregation. In the
centralized method, every sensor node must send its readings to the client host.
Suppose the depth of a sensor node is n, and it needs n ??’ 1 communications
to transmit the data produced by the sensor node to the root of the routing
tree. Figure 14(a) illustrates the centralized aggregate approach. Each node
is labelled by its distance from the root of the routing tree.
Pages:
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502