Hardware Reference
In-Depth Information
6
DYNAMIC PRIORITY SERVERS
6.1
INTRODUCTION
In this chapter
1
we discuss the problem of scheduling soft aperiodic tasks and hard
periodic tasks under dynamic priority assignments. In particular, different service
methods are introduced, the objective of which is to reduce the average response time
of aperiodic requests without compromising the schedulability of hard periodic tasks.
Periodic tasks are scheduled by the Earliest Deadline First (EDF) algorithm.
With respect to fixed-priority assignments, dynamic scheduling algorithms are char-
acterized by higher schedulability bounds, which allow the processor to be better
utilized, increase the size of aperiodic servers, and enhance aperiodic responsive-
ness. Consider, for example, a set of two periodic tasks with the same utilization
U
1
=
U
2
=0
.
3, so that
U
p
=0
.
6. If priorities are assigned to periodic tasks
based on RM and aperiodic requests are served by a Sporadic Server, the maximum
server size that guarantees periodic schedulability is given by Equation (5.24) and
is
U
max
SS
=2
/P
−
1, where
P
=(
U
1
+1)(
U
2
+1) = 1
.
69.
Hence, we have
U
max
SS
0
.
18. On the other hand, if periodic tasks are scheduled by EDF, the proces-
sor utilization bound goes up to 1
.
0, so the maximum server size can be increased up
to
U
s
=1
−
U
p
=0
.
4.
For the sake of clarity, all properties of the algorithms presented in this chapter are
proven under the following assumptions:
1
Part of this chapter is taken from the paper “Scheduling Aperiodic Tasks in Dynamic Priority Systems”
by M. Spuri and G. Buttazzo, published in
Real-Time Systems
, 10(2), March 1996.
Search WWH ::
Custom Search