A Novel Partitioning Approach for Real-Time Scheduling of
Mixed-Criticality Systems
Hayfa Ben Abdallah
1
, Hamza Gharsellaoui
1
and Sadok Bouamama
1,2
1
National School of Computer Science (ENSI), University of Manouba, Tunisia
2
Higher Colleges of Technology (HCT), Dubai, U.A.E.
Keywords:
Real-Time System RTSys, Homogeneous Multi-Core Architecture, Tasks Allocation, Clustering,
Communication Cost.
Abstract:
In real-time system (RTSys), a program is split into small tasks and distributed among several computing
elements to minimize the overall system cost. Intrinsically, tasks allocation problem is NP- hard. To overcome
this issue, it is necessary to introduce heuristics for generating near optimal solution to the given problem. This
paper deals with the problem of dependent and periodic tasks to be assigned to different cores interconnected
by a network-on-chip (NoC) in such a way that the load on each Core is almost acceptable. Further, the
development of an effective algorithm for allocating ‘N’ tasks to ‘P’ cores. The system using task clustering
to reduce the Communication Cost on the NoC. Experiment results and simulations demonstrate the efficiency
of the proposed approach.
1 INTRODUCTION
The design of embedded real-time systems is devel-
oping more and more with the increasing integra-
tion of critical functionalities for surveillance appli-
cations, particularly in the biomedical, environmen-
tal, aeronautics, mobile communication systems and
so on (Houssein and Hadi, 2016), (S. Meskina and
Z. Li, 2017). The development of these systems must
meet various challenges in terms of minimizing en-
ergy consumption. Managing such fully autonomous
embedded devices requires however to solve various
problems related to the quantity of energy available
in the battery, to the real-time scheduling of the tasks
that must be executed before their deadlines, recon-
figuration scenarios, particularly in the case of adding
tasks and the communication constraint to be able to
ensure the exchange of messages between the proces-
sors, while maintaining an acceptable level of service
quality to the system treatment.
To deal with this problem, we propose in this
work a strategy for placement and scheduling of tasks
allowing the execution of real-time applications on
an architecture containing homogeneous cores (Gam-
moudi and Chillet, 2015), a periodic tasks τ
i
is char-
acterized by two parameters (Liu and Layland, 2016),
its WCET W
i
and its period T
i
, which are assumed as
constants for all task instances (homogeneous multi-
core architecture). So to optimize one or more mea-
sure(s) of effectiveness: acceptable load on the cores
of processors, minimization of cost communication,
maximization of system reliability, etc.
A real-time system, denoted RTSys, can be imple-
mented by N dependent and periodic tasks to be as-
signed by the allocation method to P cores connected
by a network on chip (NoC) (Hu and Marculescu,
2005).
According to the requested quality of service
(QoS), the tasks may increase or decrease their exe-
cution rate to fit the requirements of other concurrent
activities, but this will not engender the systems fail-
ure.
The heuristic-based scheduling techniques are the
most common approaches for task scheduling. These
are usually classified into three classes,
* Priority-based scheduling
* Duplication-based scheduling
* Cluster-based scheduling
In priority-based scheduling, priorities are calcu-
lated and assigned to the tasks which are then sched-
uled on the processors according to their priorities.
In duplication based scheduling, while tasks are allo-
cated to a processor, its parent (and predecessor) tasks
are duplicated to occupy the idle times of the proces-
sor to eliminate the communication delay that occurs
882
Ben Abdallah, H., Gharsellaoui, H. and Bouamama, S.
A Novel Partitioning Approach for Real-Time Scheduling of Mixed-Criticality Systems.
DOI: 10.5220/0012411200003636
Paper published under CC license (CC BY-NC-ND 4.0)
In Proceedings of the 16th International Conference on Agents and Artificial Intelligence (ICAART 2024) - Volume 3, pages 882-889
ISBN: 978-989-758-680-4; ISSN: 2184-433X
Proceedings Copyright © 2024 by SCITEPRESS – Science and Technology Publications, Lda.

when message is passed from the parent tasks to the
allotted task.
In cluster-based scheduling, some tasks, that need
to communicate among themselves, are grouped to-
gether to form a cluster. Clusters are then scheduled
on to an available processor. The main problem arises
when the number of clusters is greater than the num-
ber of processors. This leads to programming the
communicating clusters on the same processor and
which remains in the nearest processor.
In this paper, a clustering based task allocation
algorithm which finds a near optimal solution to the
problem is adopted.
The remainder of this article is structured as fol-
lows: Section 2 deals with real-time systems and
more specifically with types and properties of tasks.
In the Section 3, the multicore architecture and its
scheduling approaches are presented. In Section 4,
we briefly presented our proposed method of tasks al-
location. Finally, Section 5, concluded the paper.
2 THE REAL-TIME
MIXED-CRITICALITY SYSTEM
Many embedded systems are real-time, i.e. they per-
form tasks under time constraints. Many definitions
have been proposed for the real-time systems notion.
The most widely used definition is that of Stankovic
et al (J.Stankovic, 1988). Thus, in this section, we
present many concepts related to the real-time sys-
tems from the literature.
2.1 Real-Time Systems Definition
The notion of real-time system RTSys is attached to
reactive computer system which must respect time
constraints. It depends not only on results of pro-
cessing carried out but also on temporal aspect "the
correctness of system depends not only on logical re-
sults of computation, but also on time at which results
are produced" (J.Stankovic, 1988). The real-time sys-
tem is typically composed of several processes (se-
quential) provided with temporal constraints. We call
this processes tasks. In real-time system to ensure
proper functioning of system, it is necessary to pro-
vide guarantees on respect time deadlines of tasks
that are scheduled. There are real-time applications in
field of aeronautics, automotive, telecommunications,
robotics, etc.
2.2 Definition of the Real-Time Systems
with Mixed Criticality
Nowadays, execution media are more efficient and
provide greater computing capacities than before.
Indeed, current platforms include several calcula-
tion units. Subsequently, material possibilities have
greatly increased and energy consumption has be-
come key problem (Burns and Davis, 2017). How-
ever, execution of application on medium is no longer
profitable because calculation unit is under-used.
It is therefore essential to take advantage of these
computing capacities and to minimize waste of en-
ergy. Then, some applications with different critical-
ity levels on same execution medium will be inte-
grated and some real-time systems with mixed criti-
cality have been proposed (Alan and Robert, 2017).
2.3 Real-Time Tasks
A real-time system is composed by a set of real-time
tasks subject to real-time constraints. Generally, a
real-time task is described by temporal properties and
three distinguished types.
2.3.1 Definition
A tasks τ
i
is a generic entity which has temporal
constraints and communication relations. It has time
properties that indicate arrival time, execute time and
critical time allowed for execution of the task τ
i
. Of-
ten, precedence relationships can be defined to indi-
cate order in which tasks are executed.
2.3.2 Real-Time Task Temporal Properties
The guarantee of real-time systems is ensured by re-
specting time constraints when executing application.
Generally, a real-time task is described by tempo-
ral properties defined as follows:
1. Ready time or release time "R
i
":
Date on which a task τ
i
can start its execution;
2. Worst Case Execution Time "W
i
":
This parameter is considered in majority of jobs as
worst case execution time of task τ
i
(Computing
Time "C
i
") on processor which is assigned;
3. Deadline "D
i
": Represents the instant at which the
execution of the task must be terminated;
4. Latence "L
i
": Represents the remaining time be-
fore occurrence of start or the Deadline;
5. Period "P
i
": Fixed Time interval between succes-
sive arrivals of task.
A Novel Partitioning Approach for Real-Time Scheduling of Mixed-Criticality Systems
883

* Wake up date: ↑,
* Date of Deadline: ↓,
The figure 1 below represents temporal properties
of tasks.
Figure 1: Temporal properties of real-time task.
2.3.3 Tasks Types
The representation of real-time applications in form
of tasks models operation of these applications. In
the latter, we distinguish tasks characterized by law
of arrival which defines their distribution on time and
dates creation of their instances. Tasks executing on
processor platform are types of recurring tasks, gener-
ally periodic, and non-recurring tasks, aperiodic tasks
or sporadic tasks.
Periodic Task. The periodic tasks τ
i
represent re-
curring tasks whose successive activations are sepa-
rated by constant period. They are generally charac-
terized by strict deadlines and constitute majority of
tasks composing real-time application.
According to (M.Alabau and Dechaize, 1991), a
periodic task τ
i
is modeled by the four temporal pa-
rameters: (R
i
, W
i
, D
i
, P
i
).
Aperiodic Task. In this case, activation date of ape-
riodic tasks τ
a
i
cannot be anticipated, since its execu-
tion is determined by occurrence of an internal event
(for example the arrival of message) or external (for
example operator requests). Such task is often char-
acterized by relative deadlines.
Then τ
a
i
is modeled by single time parameter: W
i
its worst case execution time.
Sporadic Task. Sporadic τ
s
i
tasks are a combina-
tion of periodic and aperiodic task, where execution
time is aperiodic but execution rate is periodic in na-
ture, these tasks are generally associated with dead-
line and represent minimum separation between two
consecutive instances.
Then τ
s
i
is modeled in general case by: (W
i
, D
i
,
T
i
min
) where T
i
min
corresponds to a minimum interval
of time separating two successive activations of spo-
radic tasks and D
i
to critical delay.
3 REAL-TIME SCHEDULING ON
MULTICORE ARCHITECTURE
This part addresses tasks scheduling on multi-core
architectures. We will first start by defining multi-
core architectures. Then, we present the different
scheduling approaches. Placing real-time task on
cores means executing it on cores. So this task and
all the other tasks that are already allocated on this
same core are schedulable of homogeneous multicore
platforms according to schedulability condition (see
equation (3)) which is specified. In this part, we study
some existing approaches in multicore architecture al-
lowing the resolution of the problems related to allo-
cation and scheduling to improve the performance of
the system and its quality of service.
In the literature, some works (Zhang, 2012)
consider that tasks scheduling in multi-core systems
is strongly dependent on tasks allocation. Jia and
al (Huang and Raabe, 2011) propose tasks alloca-
tion heuristics based on the list by assigning each of
them a priority. This list (there is only one list for
all the tasks: global approach) determines the order
of tasks allocation on different cores of processors.
At each step of the heuristic, a task is selected from
the list of candidate tasks to be scheduled on core,
then the list is updated, this is repeated until the list
is empty. This single scheduling strategy is applied to
all cores. In other words if dependent tasks are placed
on different cores then, the communication cost and
the energy consumption are so high. So, the system
performance decreases and we obtain a poor quality
of service. Indeed; it is necessary to find the best op-
timal allocation of tasks for scheduling which min-
imizes the cost of the communication. In our work
and based on (Bhardwajn and Kumar, 2013)) works,
we propose an heuristic approach to frequently place
each Cluster (contains communicating tasks) on the
same core (each C
k
core is associated with a subset
of tasks: hybrid approach) or on the nearest core to
reduce the cost of communication and adding the ver-
ification constraints on core loads (If more than one
Core, the least loaded one be chosen). The load of the
existing cores is an additional parameter L proposed
to improve the system’s performance, feasibility and
quality of services.
ICAART 2024 - 16th International Conference on Agents and Artificial Intelligence
884

3.1 Architecture of Multicore Processor
A multi-core processor contains multiple computing
units close to each other, they can communicate with
each other at high speed through bus.
Generally, increase the number of cores in pro-
cessor has advantage to execute several tasks at same
time, but doing with less energy than other architec-
tures. So multi-core allows more performance with
several clocked cores at lower frequency, thus offer-
ing more flexibility and energy savings.
3.2 Multicore Scheduling
Scheduling is a mechanism by which cores have ac-
cess to place or run in order a set of tasks for ap-
plication defines. The main goal is properly bal-
ance use of cores so that tasks are execute out most
efficiently. Then we will talk about task sched-
uler (Joel Goossens, 2016).
3.3 Multicore Scheduling Approach
Multi-core scheduling is a 2 dimensional problem.
First, allocation problem (spatial organization) deter-
mines for each tasks which core should run on. Sec-
ond, scheduling problem (temporal organization) de-
fines date and order execution of tasks.
There are two basic types of multi-core strategies:
global approaches and partitioning approaches.
3.3.1 Global Approach
The objective is to allocate at each instant tasks to
cores P. For this, there is a single queue for all tasks
and single scheduling strategy is used which applies
the set of cores (see Figure 2). In this case, an essen-
tial property is that tasks migration is allowed.
τ
1
τ
2
τ
n
C
1
C
2
C
p
global scheduler
Figure 2: Global approach.
3.3.2 Partitioning Approach
The objective is to definitively assign each task to one
core or to define subset of tasks each assigned to core.
It is thus possible to apply single-core strategies
on each of the architecture’s processors, in which each
processor has their own execution queue (see Figure
3).
τ
1
τ
2
τ
n
allocation algorithm
local scheduler1 local scheduler2 local scheduler n
C
1
C
2
C
p
τ
1
,τ
5
τ
2
,τ
3
τ
4
Figure 3: Partitioning Approach.
3.3.3 Hybrid Approach
The hybrid or semi-partitioned approach is derived
from the partitioned approach. In this approach, cer-
tain occurrences of tasks can be executed on different
cores (migration).
Each core receives certain tasks after the partition-
ing heuristic and can perform an exchange of tasks
with other cores, this is called task migration.
It should be noted that for all these approaches,
there is a necessary condition of schedulability, (see
equation (3)).
In our work, we propose a development strategy
based on an hybrid approach because the latter re-
duces problems of communication cost and energy
consumption. So, this approach makes a feasible sys-
tem with good quality of services.
3.4 Utilization Factor Model
Processor / Core
According to Liu and Layland (Liu and Layland,
2016), the rate of the processor utilization by a task
τ
i
is:
υ
τ
i
=
w
i
T
i
(1)
Thus, the utilization factor (or processor utilization)
is a percentage of time that processor passes to exe-
cute all the N tasks that composed the system, and is
A Novel Partitioning Approach for Real-Time Scheduling of Mixed-Criticality Systems
885

worth:
U
c
k
=
∑
N
i=1
υ
τ
i
=
∑
N
i=1
w
i
T
i
, ∀k ∈ [1...P].
(2)
A set of N periodic tasks on P cores can be scheduled
if the previous condition is valid. According to (Jin
and Schiavone, 2007),
U
c
k
≤ 1
(3)
The equation (3) is a real-time constraint noted later
by RTConst which defines the necessary condition for
schedulability of N tasks on P Cores.
3.5 Communication Model
Since tasks are dependent and can exchange messages
between them, we denote by M
i, j
the periodic mes-
sage sent from τ
i
to τ
j
.
Based on the work of Bach and al (Bach Duy Bui
and Caccamo, 2012), each message M
i, j
is character-
ized by:
* A period TM
i, j
: The regular time of arrival,
* Time required to send message WM
i, j
,
* Deadline DM
i, j
,
* This is absolute limit time not be exceeded,
* Size SM
i, j
: Relative size of message in bytes,
The Communication between two cores is ensured by
a communication medium which can be bus, NoC,
etc.
The network-on-chip (NoC) architecture implies
a communication cost between each pair of cores C
k
and C
l
depending on the distance between them. We
denote by Cost
C
k
,C
l
this cost.
We model each NoC architecture by a specific cost
matrix Cost such as
Cost
C
k
,C
l
=
(
X
C
k
,C
l
i f k ̸= l, ∀k, l ∈ [1...P].
0 otherwise.
where X
C
k
,C
l
is a constant giving the cost of transfer-
ring one byte from core C
k
to core C
l
.
Similar to works (Huang and Raabe, 2011), com-
munication Cost between each pair of cores C
k
and C
l
,
it is necessary to multiply the volume of exchanged
data by Cost
C
k
,C
l
. Then, the total cost of communica-
tions is given by:
TotalCost =
∑
P
k=1
∑
P
l=1
∑
N
i=1
∑
N
j=1
Cost
C
k
,C
l
∗ SM
i, j
(4)
4 Proposed Tasks Allocation Method
We present an allocation strategy of tasks on one core
based on Clustering concept which groups communi-
cating tasks on the same or the nearest cores. This
core must check real-time constraints RTConst (see
equation (3)).
4.0.1 Architecture Model
Based on (Bhardwajn and Kumar, 2013) work, we as-
sume that each core C
k
(k ∈ [1 . . p]) schedules locally
assigned tasks with EDF algorithm and increases the
number of tasks to be executed. Given that we are
considering a multicore platform, we assume also that
each task τ
i
can be executed on any less loaded core.
In case of homogeneous multicore architecture
(identical cores), we not need of allocation matrix for
tasks because can be execute of any cores who checks
real-time constraint RTConst.
So, to verify that all tasks are allocated on any
cores. We define a condition of tasks allocation Z on
cores :
Z
i,k
= task τ
i
is allocated on core C
k
.
∑
P
k=1
Z
i,k
= 1, ∀i ∈ [1...N]
(5)
4.0.2 Strategy Principle
The proposed strategy solves tasks scheduling prob-
lem. The idea is to group communicating tasks into
different subsets called Cluster (denoted CL) while
checking the feasibility constraint at the same time.
If there is more than one core, the least loaded one
will be chosen to improve the system performance
and quality of services. we allocate tasks that com-
municate frequently on the same or the nearest cores
following the equation (3). The objective of the pro-
posed strategy is to ensure these following points:
1. Each task must be allocated at only on the core
that can execute it,
2. The load of each core must not exceed its maxi-
mum load (U
c
k
≤ 1, ∀k ∈ [1... p]),
3. The period of each task must not exceed its maxi-
mum period (T
i
max
),
Before describing the strategy steps, it is neces-
sary to define some preliminaries to be used later.
* Cluster (CL).
Group of tasks that exchange data with each other.
Example: τ
1
, τ
2
and τ
3
are grouped in CL
1
, (See
figure 4).
* Core Processor (CP).
Is the core which can execute all the tasks in the
same Cluster such that its utilization rate is less
than or equal to 1.
Example: Looking at Figure 4, we notice that core
C
1
can execute all tasks of Cluster CL
1
. So, if its
utilization rate (supporting all these tasks) is less
than or equal to 1, then it is a CP.
* CPCL.
Cluster CL which can be schedulable with one or
ICAART 2024 - 16th International Conference on Agents and Artificial Intelligence
886

Table 1: Comparison of communication cost related to tasks increasing.
tasks N Cores P Alg.(Bhardwajn and Kumar, 2013) Alg.(Govil, 2011) Our.Alg
6 4 106 101 90
7 4 140 135 112
8 4 188 169 135
9 4 243 224 201
10 4 259 251 222
11 4 333 319 290
12 4 378 365 345
13 4 435 401 375
14 4 480 465 450
15 4 518 502 472
16 4 533 522 500
17 4 565 545 523
18 4 588 576 552
19 4 617 591 550
20 4 652 620 592
Total 6035 5786 5409
Figure 4: Example of cluster CL.
more Core Processors CP,
Example: If the cluster CL
1
can be scheduled on
C
1
(U
C
1
≤ 1), so it is called CP and it becomes a
CPCL (see figure 4).
* Traffic Data (TD).
Amount of data transferred (in bytes) of Cluster
CL in unit time (ut), (see Figure 4).
Example:
T D(CL
1
) = ut ∗ (
SM
1,3
+SM
1,2
T
1
) and T D(CL
2
) = 0.
(6)
5 SIMULATION RESULTS
A system RTSys, can be implemented by N dependent
and periodic tasks to be assigned to P cores linked by
a network-on-chip (NoC) in order to ensure the com-
munication between dependent tasks. Reducing the
cost of communications between cores becomes a ma-
jor concern to ensure high-performance and reliability
of such systems.
To show efficiency of proposed strategy, we car-
ried comparisons with allocation algorithms of Poorn-
ima and al (Bhardwajn and Kumar, 2013), kapil
(Govil, 2011). This section therefore makes possi-
ble to situate the proposed strategy in relation to al-
gorithms mentioned above. We propose to implement
the three approaches using the same case study pre-
sented in section (Bhardwajn and Kumar, 2013).
The objective is to allocate tasks on different cores
and calculate the total communication cost of each ap-
proach. The 1-D Mesh network (Konstantakopoulos,
2007) communication medium is used to connect the
presented cores in the previous example.
5.1 Comparison of Algorithms by
Increasing the Number of Tasks
First, the comparison is performed by fixing the num-
ber of cores to 4 with a variation number of tasks. So,
the table 1 represents the communication cost for each
algorithm.
Experimental results on our strategy show that it
offers 10.5% (
6035−5409
6035
∗ 100) communication cost
profit compared to algorithm proposed by poornima
and al. (Bhardwajn and Kumar, 2013). It offers
6.5% (
5786−5409
5786
∗ 100) profit compared to algorithm
proposed by kapil. (Govil, 2011).
So, Kapil algorithm (Govil, 2011) is an im-
proved extension of the one proposed by poornima
and al. (Bhardwajn and Kumar, 2013).
A Novel Partitioning Approach for Real-Time Scheduling of Mixed-Criticality Systems
887

6
8 10 12 14
16
18 20
200
400
600
Tasks→
Cost in Units →
Poornima and al.algorithm
Kapil Govil.algorithm
Proposed algorithm
Figure 5: Cost of communication when tasks are in an increasing order and number of cores is 4.
Table 2: Comparison of communication cost when number of cores is in increasing order.
Cores P Tasks N Alg.(Bhardwajn and Kumar, 2013) Alg.(Govil, 2011) Our. Alg
4 50 645 620 608
5 50 672 640 630
6 50 669 640 625
7 50 678 650 638
8 50 675 649 636
9 50 672 647 635
10 50 682 655 646
11 50 688 666 658
12 50 685 665 659
13 50 685 660 655
14 50 678 658 650
15 50 675 655 645
Total 8104 7805 7685
4
6
8 10 12 14
16
600
620
640
660
680
Cores→
Cost in Units →
Poornima and al.algorithm
Kapil Govil.algorithm
Proposed algorithm
Figure 6: Cost of communication when cores are in an increasing order and number of tasks is 50.
ICAART 2024 - 16th International Conference on Agents and Artificial Intelligence
888

Figure 5 presents curves which compare commu-
nication cost in each approach.
5.2 Comparison of Algorithms by
Increasing the Number of Cores
Then, we increase the number of cores by fixing the
number of tasks to 50. Thus, the table 2 represents the
communication cost for each algorithm.
When measuring the profit obtained in terms of
communication cost, we find that our strategy of-
fers a little more than 5% (
8104−7685
8104
∗ 100) of profit
compared to poornima algorithm and 2% (
7805−7685
7805
∗
100) profit compared to that proposed by kapil. Fig-
ure 6 presents curves which compare communication
cost in each approach.
The simulations show that our strategy is better
than those proposed in (Bhardwajn and Kumar, 2013)
and (Govil, 2011). This is explained by efficiency of
strategy since it offers several heuristics which gives
more flexibility to placement level.
5.3 Evaluation
It can be observed from Figure 5 that the values of
the total optimal cost obtained by the present algo-
rithm are better compared to those obtained in (Govil,
2011) and (Bhardwajn and Kumar, 2013), in the case,
when the number of cores is kept fixed and the num-
ber of tasks is taken in an increasing order. The simi-
lar observation can also be made from Figure 6 in the
case when the number of tasks is fixed and the number
of cores is taken in increasing order.
Thus, it is concluded that the present algorithm
results have better optimal cost in both cases.
6 CONCLUSION
In this paper, the problem of periodic tasks allocation
on a homogeneous multicore architecture using tasks
clustering, is discussed. As the task allocation prob-
lem is known to be NP-hard.
Our strategy proposes an allocation of tasks which
reduces cost of communication between cores and
also suggests reducing distance between tasks if the
allocation on same core is not possible to obtain the
system feasibility.
From the experimental results, we conclude that
the proposed solution improves the cost communica-
tion in the whole system while keeping its feasibility.
REFERENCES
Alan, B. and Robert, D. (Jan 2017). Mixed critical-
ity systems-a review. Department of Computer Sci-
ence, University of York, Technical. Report, pages
1–69. http://www-users.cs.york.ac.uk/burns/
review.pdf.
Bach Duy Bui, R. P. and Caccamo, M. (2012). Real-
time scheduling of concurrent transactions in multido-
main ring buses. IEEE Transactions on Computers,
61(9):1311–1324.
Bhardwajn, P. and Kumar, V. (2013). An effective load bal-
ancing task allocation algorithm using task clustering.
International Journal of Computer Applications.
Burns, A. and Davis, R. (November 2017). A survey of
research into mixed criticality systems. ACM Comput,
82(37):37–50.
Gammoudi, A. Benzina, M. K. and Chillet, D. (2015).
New pack oriented solutions for energy-aware feasi-
ble adaptive real-time systems. Intell. Softw. Method-
ologies Tools Techn, page 73–86.
Govil, K. (2011). A smart algorithm for dynamic task al-
location for distributed processing environment. In-
ternational Journal of Computer Applications, page
13–19.
Houssein, H. E. and Hadi, M. A. E. (2016). Energy effi-
cient scheduler of aperiodic jobs for real-time embed-
ded systems. In J. Autom. Comput, page 1–11.
Hu, J. and Marculescu, R. (2005). Energy-and
performance-aware mapping for regular noc archi-
tectures. IEEE Trans. Comput.-Aided Design In-
tegr.Circuits Syst, 24(4):551–562.
Huang, J. and Raabe, A. (2011). Energy-aware task allo-
cation for network-on-chip based heterogeneous mul-
tiprocessor systems. In Parallel, Distributed and
Network-Based Processing (PDP), 2011 19th Eu-
romicro International Conference on, page 447–454.
IEEE.
Jin, S. and Schiavone, G. (2007). A performance study of
many cores task scheduling algorithms. London, 2nd
edition.
Joel Goossens, E. G. a. L. C.-G. (November 2016). Peri-
odicity of real-time schedules for dependent periodic
tasks on identical multiprocessor platforms. Real-
Time Systems, 52(6):808–832.
J.Stankovic (1988). Misconceptins about real-time comput-
ing. IEEE Computer, London, 2nd edition.
Konstantakopoulos, T. K. (2007). Energy scalability of
on-chip interconnection networks. PhD thesis, Mas-
sachusetts Institute of Technology.
Liu, C. L. and Layland, J. W. (2016). Scheduling algo-
rithms for multiprogramming in a hard-real-time en-
vironment. volume 12, page 101–111.
M.Alabau and Dechaize, T. (1991). Ordonnancement
temps-réel par échéance. London, 2nd edition.
S. Meskina, N. Doggaz, M. K. and Z. Li, M. f. (2017). for
smart grids recovery. EEE Trans. Syst. Man. Cybern.
Syst, 47(7):1284–1300.
Zhang, H. (2012). ordonnancement de tâches temps réel
dans les systèmes multicoeur. PhD thesis, Université
de Nantes.
A Novel Partitioning Approach for Real-Time Scheduling of Mixed-Criticality Systems
889