Following are some of the important design issues in such sinkinitiated
querying sensor networks.
18
Addressing Requirements
Since the sink is interested in disseminating queries to a subset of sensor nodes,
a node-addressing strategy should be in place. It may not be necessary for the
sink to be aware of addresses of all the nodes in the network. In most cases,
it is su?±cient if the node-addressing is tied to the node locations, because
typically the sink is interested in querying a subset of nodes located in a
specific region.
Selective Querying
In order that the sink be able to query a subset of sensor nodes selectively,
it is also necessary to have a routing infrastructure that allows for limited
broadcasting in the selected region [56, 11]. For this, a routing broadcast must
use information about node locations. Hence it is necessary to have a reliable
transport protocol such as PSFQ [54] for dissemination of control messages
from the sink to the sensor nodes. There is also a need for performing data
aggregation within the selected region to avoid sending the raw data to the
sink. This could be achieved via clustering and/or hop-by-hop aggregation.
Querying Language
Since querying of sensor nodes for data is identical to the paradigm of querying
in databases, selective querying sche-mes use an SQL-based querying language
for query-based communication between the sink and the sensor nodes. Such
a querying language enables the sink to advertise its interest using the meta
attributes of data such as FROM, WHERE, SAMPLE PERIOD, HAVING AVG(),
HAVING MAX(), etc.
Pages:
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67