In DIM, nodes can have
di?®erently-sized zones (as Figure 10 illustrates), and each node automatically
discovers its zone codes. The zone tree corresponding to Figure 10 is shown
in Figure 11.
0 1
0 1
0 1
0 1
0 1 0 1
0 1
Fig. 11. Zone tree of sensor network [24].
Let us discuss how events are hashed to zones. Consider a DIM that aims
to support m distinct attributes. Let us denote these attributes Al, ...,Am.
For ease of exposition, assume that the depth of every zone in the network
is k, m is a sub-multiple of k, and that this value of k is known to every
node. Furthermore, assume that all attribute values have been normalized to
be between 0 and 1. The DIM hashing scheme assigns a k bit zone code to an
event as follows. For 0 ?· i ?· m, if Ai < 0.5, the i-th bit of the zone code is
assigned 0, else 1. For m+1 ?· i ?· 2m, if Ai??’m < 0.25 or 0.5 ?· Ai??’m < 0.75,
the i-th bit of the zone is assigned 0, else 1. This procedure repeats until all k
bits have been assigned. As an example, consider event < 0.3, 0.8 >. For this
event, the 5-bit zone code is 01110.
At least conceptually, the rest of the DIM system is relatively easy to
describe. To insert an event e, DIM computes the zone code of e. GPSR is
then used to route e to zone Z whose zone code has the longest matching
prefix of the e??™s zone code. Since e??™s zone code can be mapped to a geographic
location, the target zone will be somewhere ???near??™ that location.
Pages:
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497