305
Jinbao Li, Zhipeng Cai, and Jianzhong Li
Apex
Clusterhead
Fig. 8. The DIMENSIONS hierarchy [16].
Similar with DIMENSIONS, DIFS uses only one key attribute of the sensor
data. Geographic hashing and spatial decomposition are used to construct
the multi-rooted hierarchical structure, which is a one-dimension index. This
index has two features. First, it builds a geographically-aligned multi-rooted
hierarchy designed to avoid the root as a bottleneck in DIMENSIONS. Second,
it e?±ciently propagates information summarizing the hierarchy that can be
used to prune traversal at higher levels of the hierarchy.
We take temperature as an example to introduce how one dimension index
is built in DIFS. Suppose the range of temperature is between 1?—¦C and 100?—¦C.
The construction process of the index is best described with reference to Figure
9. As with DIMENSIONS, DIFS divides the geographical area encompassed
by the sensor network equally, which constructs the d-level hierarchical tree.
The di?®erence from DIMENSIONS is that the tree has multiple roots instead
of one. Now, suppose that an event with a temperature value 26 is generated
at some node within the network. It is stored in a logical leaf node of the DIFS
hierarchy as follows. (1) Hash a unique name for the index (say ???temperature??™)
together with a textual representation of the range of values (???0 : 100??™ in our
example). (2) The logical leaf node of the DIFS is obtained by the hashing in
step (1).
Pages:
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492