Database Reference
In-Depth Information
evaluates this query by interpolating the value of temperature at fixed
intervals on the x- and y-axis; this is similar to database view material-
ization [19]. Then the positions (
x,y
) where the interpolated tempera-
ture value is greater than 10
◦
C are returned.
Admittedly, although updating the model-based view is ecient, but
for processing queries the model-based view should be materialized at
a certain fixed set of points. This procedure produces a large amount
of overhead when the number of independent variables is large, since it
dramatically increases the number of points where the view should be
materialized.
SELECT
x,y
FROM
RegModel
WHERE
v>
10
◦
C
Query 2.5: Querying model-based views.
4.3 Symbolic Query Evaluation
This approach is proposed by the FunctionDB [64] system. Func-
tionDB, like MauveDB, also interpolates the values of the dependent
variable, and then uses the interpolated values for query processing.
As discussed before, the main problem with value interpolation is that
the number of points, where the sensor values should be interpolated,
increase dramatically as a function of the number of independent vari-
ables. As a solution to this problem, FunctionDB symbolically executes
the filter (for example, the
WHERE
clause in Query 2.5) and obtains feasi-
ble regions of the independent variables. These feasible regions are the
regions that include the exact response to the query, at the same time
contain a significantly low number of values to interpolate. FunctionDB
model-based
views
50
50
20
20
10
10
model-based
views are
continuously
updated
10
10
20
20
40
40
time
t
1
t
2
v
22
v
13
v
11
s
2
s
2
s
3
s
3
s
1
s
1
-- sensors
-- sensor values
Figure 2.9.
Example of the
RegModel
view with three sensors.
RegModel
is incre-
mentally updated as new sensor values are acquired.
