Hardware Reference
In-Depth Information
v(f )
i
v(f )
i
soft
non real−time
f
i
f
i
d
i
v(f )
i
v(f )
i
on−time
firm
f
i
f
i
d
d
i
i
Figure 2.17
Example of cost functions for different types of tasks.
In other cases (d), executing a task after its deadline does not cause catastrophic con-
sequences, but there is no benefit for the system, thus the utility function is zero after
the deadline.
When utility functions are defined on the tasks, the performance of a scheduling al-
gorithm can be measured by the
cumulative value
, given by the sum of the utility
functions computed at each completion time:
n
Cumulative value
=
v
(
f
i
)
.
i
=1
This type of metrics is very useful for evaluating the performance of a system during
overload conditions, and it is considered in more detail in Chapter 9.
2.4
SCHEDULING ANOMALIES
In this section we describe some singular examples that clearly illustrate that real-
time computing is not equivalent to fast computing, since, for example, an increase of
computational power in the supporting hardware does not always cause an improve-
ment of performance. These particular situations, called Richard's anomalies, were
described by Graham in 1976 and refer to task sets with precedence relations executed
in a multiprocessor environment.
Search WWH ::
Custom Search