Database Reference
In-Depth Information
SELECT spatial granule, AVG(temp)
FROM data s [Range By 5 min]
(
SELECT spatial granule, avg(temp) as avg,
stdev(temp) as stdev
FROM data [Range By 5 min]) as a
WHERE a.spatial granule = s.spatial granule
AND a.avg + (2*a.stdev) < s.temp
AND a.avg - (2*a.stdev) > s.temp
Query 2.2: An example of anomaly detection using a SQL statement.
work that enables data owners to specify or discover groups of attributes
that are correlated, and apply statistical methods that validate and clean
the sensor values using such dependencies.
4. Model-Based Query Processing
In this section we elaborate another important task in sensor data
management - query processing. We primarily focus on in-network and
centralized query processing approaches. We consider different queries
assuming the sensor network described in Section 2.1, and then discuss
how each approach processes these queries. In Section 2, however, we
followed an approach where we chose a singe query (i.e., Query 2.1)
and demonstrated how different techniques processed this query. On the
contrary, in this section, we chose different queries for all the approaches,
and then discuss these approaches along with the queries. We follow this
procedure since, unlike Section 2, the assumptions made by each query
processing technique are different. Thus, for highlighting the impact
of these assumptions and simplifying the discussion, we select different
queries for each approach.
4.1 In-Network Query Processing
In-network query processing first builds an overlay network (like, the
SRT discussed in Section 2.3.1). Then, the overlay network is used
for increasing the eciency of aggregating sensor values and processing
queries. For instance, while processing a threshold query, parent nodes
send the query to the child nodes only when the query threshold con-
dition overlaps with the range of sensor values contained in the child
nodes, which is stored in the parent node's local memory.
Consider the threshold query given in Query 2.3. Query 2.3 requests
the sensor identifiers of all the sensors that have sensed a temperature
greater than 10
◦
C at the current time instance. Before answering this
query, we assume that we have already constructed a SRT as described
