Information Technology Reference
In-Depth Information
two virtual processors VP
ðw
h
Þ
1
and VP
ðw
h
Þ
2
with different utilization factors UVP
ðw
h
Þ
1
and UVP
ðw
h
2
to adapt the system to its environment with a minimum response
times. UVP
ðw
h
1
corresponds to the processor utilization of the system before any
addition scenario, and UVP
ðw
h
Þ
2
can be assigned to any value lower than 1 according
to user requirements.
Therefore, based on the research work in Buttazzo and Stankovic (
1993
) which
provides a window-constrained-based method to determine how much a task can
increase its computation time without missing its deadline under EDF scheduling
[for more informations about the window-constrained-based method, you can see
Buttazzo and Stankovic (
1993
)]. We propose in our thesis work, a window con-
strained schedule which is used to separate old and new tasks. Old and new tasks
are located in different windows to schedule the system with a minimum response
times. Idle periods in VP
ðw
h
1
, which appear alternatively with busy periods, are
considered as logical windows for the execution of the second virtual processor
VP
ðw
h
Þ
2
first logical window corresponding to VP
ðw
h
Þ
1
. In this case, a
is reserved for
old tasks (periodic and sporadic) that should be recon
gured to meet their deadlines
and to reduce their response time, and a second window corresponding to VP
ðw
h
2
is
reserved for new sporadic tasks with an optimal response time after any sporadic
tasks addition. The physical processor of
n
ðw
h
Þ
will be running with this solution in
two phases for the minimization of response times.
We assume in the following that new sporadic tasks are dynamically added to a
system and request the processor at a time t which is not smaller than Pi
i
(=D
i
). After
any recon
guration scenario
w
h
and in order to keep only two virtual processors in
n
ðw
h
Þ
, the proposed intelligent agent automatically merges VP
ðw
h
1
Þ
1
the system
and
VP
ðw
h
1
Þ
2
into VP
ðw
h
Þ
1
and creates also a new VP
2
for the recon
guration scenario
w
h
named VP
ðw
h
Þ
2
, to adapt old and new tasks, respectively. The VP
ðw
h
Þ
2
is assumed to be
a located logical pool in idle periods of VP
ðw
h
1
and used to execute new added tasks.
The old tasks are assumed to be executed by the
first virtual processor VP
ðw
h
Þ
1
.
Running Example
To illustrate the key point of the proposed recon
guration approach, we consider
the Volvo task system shown in Tables
1
and
2
, as a motivational example noted
n
composed of 2 characterized periodic tasks (
s
A
and
s
B
) and 3 sporadic tasks (
r
D
,
r
E
, and
r
H
)asa
rst task set (
1
) and 3 added sporadic tasks as a second task set to
be added later (
2
).
For example, the processor utilization factor of
¼
P
i¼1
C
i
n
in Table
1
is U
T
i
or D
i
,
C
A
C
D
B
þ
C
D
D
þ
C
D
E
þ
C
D
H
¼
(D
i
,
is in the case of sporadic tasks)
¼
T
A
þ
0
:
2
þ
0
:
4
þ
0
:
12
þ
0
:
08
þ
0
:
004
¼
0
:
804
1, so the system
n
is feasible.
Search WWH ::
Custom Search