Formal Behavioral Modeling of Real-time Operating Systems
C´edrick Lelionnais
1
, Matthias Brun
1
, J´erˆome Delatour
1
, Olivier H. Roux
2
and Charlotte Seidner
2
1
ESEO-TRAME, 4 Rue Merlet de la Boulaye, B.P.30926, 49009 Angers cedex 01, France
2
LUNAM Universit´e, IRCCyN-Ecole Centrale de Nantes, 1 Rue de la Noe, B.P.92101, 44321 Nantes cedex 03, France
Keywords:
Model Driven Engineering, Real-time Operating Systems, Behavioral Modeling, Formalization, Time Petri
Nets, Application Deployment.
Abstract:
Faced with the growing problems of complexity, heterogeneity and upgradability of Real-Time Embedded
Systems (RTESs), model-based frameworks dedicated to the application deployments facilitate the design and
the development of such systems. Within these frameworks, taking into account the Real-Time Operating
Systems (RTOSs) has become essential. These frameworks include transformation tools able to generate a
code that is portable to the specified RTOS. Moreover, certain tools can generate formal models that are used
for the verification and validation of the RTESs. However, the RTOSs technological concepts are considered
in an implicit way, which involves a lack of genericity of the transformations. Some works have focused on
the explicit description of the RTOSs. Such a description offers the possibility to take into account a model
entirely dedicated to a targeted RTOS as a parameter of the transformation. Nevertheless, this method does not
allow to verify the expected properties on the application, since the RTOSs behavior is not observable. The
methodology presented in this paper tends to explicitly consider the formal description of the RTOSs behavior
during an application deployment. This approach aims both at making each transformation generic and at
verifying the deployment correctness.
1 INTRODUCTION
Nowadays, Real-Time Embedded Systems (RTESs)
increasingly surround us in various domains (air-
crafts, automotive, cell phones, robotics...). RTES
engineers are confronted with the challenge of de-
veloping more complex, higher quality systems, with
shorter development cycles at lower costs. Within this
context, reuse, maintainability and portability have
become major issues in RTES design processes.
From this point of view, some frameworks have
been developed with the aim of generating code.
Starting from a detailed model written in such a lan-
guage as AADL (Society of Automotive Engineer
(SAE), 2004) or UML (Object Management Group
(OMG), 2007b), the specific code for the deployment
of an application on a Real-Time Operating System
(RTOS) is then generated. These code generating
frameworks are generally used after some Verification
and Validation (V&V) activities have been performed
on the models. In this way, some research works have
dealt with the transformation of those models into for-
mal models, using such formalisms as Time Petri Nets
(TPNs) or Finite State Machines (FSMs)...) as prese-
nted for instance in (Renault et al., 2009) (Berthomieu
et al., 2010).
However, the correctness of the application de-
ployed on the RTOS is not verified during these V&V
activities. It is therefore difficult to ensure that the
generated code effectively corresponds to what was
designed. Moreover the consideration of the RTOSs
is often embeded within the processus of generation
(or seen in simplified terms). This approach is then far
from flexible since each translation must be adapted in
accordance with the targeted RTOS. Due to the signif-
icant number of existing RTOSs, this results in having
to maintain a large amount of translations.
In consequence, this paper relates to the following
issues. Which methodology should be adopted in or-
der to both deploy an application on a targeted RTOS
and check the correctness of such a deployment? Is
it possible to consider any RTOS during the deploy-
ment? (and if so, how?) How can the behavior of
any RTOS be described within such frameworks? To
bring an answer to these questions, a method has been
experimented to take these factors into account within
a single framework. The basic idea is to add the for-
mal description of any RTOS. Each description can
407
Lelionnais C., Brun M., Delatour J., H. Roux O. and Seidner C..
Formal Behavioral Modeling of Real-time Operating Systems.
DOI: 10.5220/0004156204070414
In Proceedings of the 14th International Conference on Enterprise Information Systems (MDDIS-2012), pages 407-414
ISBN: 978-989-8565-11-2
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)
then be considered as a parameter of the transforma-
tion, so as to keep the framework generic. Lastly, this
formalization should let V&V tools able to verify the
correctness of such deployments.
This paper is divided into the following sections.
Section 2 presents some works in conjunction with
the issues described above. As a result of that re-
search, the approach adopted is developed in Sec-
tion 3. It has proven the need to consider the behav-
ior of RTOSs during a deployment of an application.
Then Section 4 presents the strategy proposed to for-
mally take into account the behavioral description of
RTOSs. Consequently the benefits and the limits of
the strategy are showed in Section 5 and discussed in
Section 6. Finally we conclude in Section 7.
2 RELATED WORK
2.1 Code Generating Frameworks
Code generating frameworks are used in a context of
application deployment. Code generation is the pro-
cess of transforming code from one representation to
another one. Often, this is from a higher level, such as
an UML diagram modeling an application, to a lower
level, such as a C code program that is portable to an
execution software platform.
From a real-time point of view, code generating
frameworks are used in a context of application de-
ployment to a specific RTOS. The information con-
cerning RTOSs are given in a program which im-
plements the deployment. For instance, Ocarina de-
scribed the information of a middleware with Ada
code in its Gaia generator (Vergnaud and Zalila,
2006).
Similarly, the Real-Time Workshop (RTW) tool
(The MathWorks, 2007) uses templates catching both
technological concepts of the RTOS and program-
ming language information. These templates struc-
ture the source code to generate.
As a result, these frameworks involve an implicit
description of the RTOSs during the processus of gen-
eration. This raises the problem of genericity of these
frameworks since each deployment corresponds to a
unique code generator.
2.2 RTOSs Modeling
In order to describe the Operating Sytems (OSs) con-
cepts, some model-based frameworks have been de-
veloped. Some works have focused on both structural
and behavioral aspects of the OSs. Revolving around
these two axes, the description of the OSs is made ex-
plicit and observable through formal models. Explicit
means that such a description is entirely dedicated to
the representation of the OS.
Metropolis (Team, 2004) uses formal languages
such as the Linear Temporal Logic. It offers the pos-
sibility to verify some properties of an application de-
ployed on a wide range of OSs.
Generic Modeling Environment (GME) (Davis,
2003) contributes to the description of OSs which
are not necessarely real-time and analysis platforms.
GME proposes assistance tools for the migration from
one OS to another. In addition, this environment of-
fers the possibility to insert formal metamodels. The
descriptions thereby made can be used by V&V tools.
Nevertheless, the genericity of the modeling lan-
guages used within these frameworks does not make
the modeling of platforms easier dedicated to a partic-
ular domain, such as the real-time domain, any easier.
More specifically to the RTESs, Ptolemy (Lee,
2003) is a framework which describes execution mod-
els. Most of these models support the actor-oriented
design focused on the concurrency and the communi-
cation between the components of a RTES. However,
the concepts of structural representation from this ap-
proach are only intended for the applications model-
ing, but not for RTESs themselves.
In spite of the fact that the OSs description is ex-
plicit, the works outlined above are less in line with
the RTESs. Furthermore, the separation of the con-
cerns does not clearly appear. This impacts on the in-
tervention of each domain specialist such as the appli-
cation or the OSs technology. The lack of genericity
previously noted arises once again.
Other researches have led to consider the de-
scription of RTOS during the application deployment.
This approach relies on a Model Driven Engineer-
ing (MDE) context. MDE promotes the distinction
of platform-specific modeling artefacts from those
that are platform-independent. The platform (RTOS)
model is considered as a parameter of this deploy-
ment. Each model is described by a modeling lan-
guage. Concerning the RTOS modeling, we can note
the Software Resource Modeling (SRM) (Thomas
et al., 2008) package which is a MARTE’s UML
profile (Object Management Group (OMG), 2007a).
With SRM, RTOSs can be modeled using stereotyped
concepts from the real-time software domain.
For another example, RTEPML (Real-Time Em-
bedded Platform Modeling Language) (Brun and De-
latour, 2011) was developped with the aim of defining
concepts dedicated to the real-time domain for mod-
eling RTOSs.
Contrary to the first noted frameworks, SRM
ICEIS2012-14thInternationalConferenceonEnterpriseInformationSystems
408
and RTEPML better separate the concerns previously
highlighted (RTOSs structure, transformations, de-
ployments choice...). Nevertheless, only the structural
aspect is taken into account with these modeling lan-
guages. The behavior resulting from the concurrent
activities of the RTESs components is not observable
in the deployed application models. As a result, no
V&V activities can be applied to check the correct-
ness of such a deployment.
3 APPROACH ADOPTED
The previous section showed the interest to well dis-
tinguish the concerns for the intervention of each spe-
cialist for real-time applications deployment. In addi-
tion, it has been raised to explicitly consider the de-
scription of the RTOSs behavior within a code gen-
erating framework. Our approach relies therefore on
the extension of a model-basedframeworkto formally
observe the behavior of a deployed application on a
RTOS. This should add other concerns such as the
RTOSs behavior and some V&V activities on the de-
ployment at design phase.
As a result, we have chosen the RTEPML frame-
work as it has been developed within our team. In
addition this modeling language already contributes
to the explicit description of the RTOSs. However,
the behavioral aspect of RTOSs can not be modeled
with RTEPML. Therefore modeling behavior with
RTEPML means extending this language. That is why
we have sought to enrich it in order to define behav-
ioral concepts of RTOSs.
3.1 RTEPML with MDE
Based on the MDE initiative, the application deploy-
ment emanates from semi-generic transformations of
models. MDE is based on an architecture (Model
Driven Architecture (MDA)) (Object Management
Group (OMG), 2001) (see Figure 1) involving a de-
sign process where a Platform-Independent Model
(PIM) (description of the application without consid-
ering the platform) is transformed into a Platform-
Specific Model (PSM) (description of the application
deployed on the platform), according to a Platform
Description Model (PDM) (description of the plat-
form). Each model conforms to a modeling language,
also named metamodel.
In accordance with Figure 1, RTEPML is a meta-
model which has been built to model PDMs. The
concepts of the targeted RTOS are then defined by
a PDM. So far, the modeling with RTEPML goes
through the structural characterizations (resources
Figure 1: MDA principle.
and services) of RTOSs. The resources represent the
concepts offered by the platform. For instance, in a
RTOS, the tasks, the semaphores, the mailboxes are
considered as resources. The services are offered by
the resources and caught at API’s level (Application
Programming Interface). For example, services asso-
ciated with semaphores are the take and release prim-
itives.
Consequently, in conjunction with this approach,
the deployment consists both in instantiating RTOS
resources and in using RTOS call services for imple-
menting the application. Then the generated PSM is
provided for a code generation process. Finally, it is
worth noting that the transformationserving to the de-
ployment is generic thanks to the explicit considera-
tion of the RTOSs.
3.2 Illustration around Semaphore
Sharing
We have mentioned the idea to extend RTEPML at the
beginning of this section. This extension should en-
able to consider the RTOS behavior during an appli-
cation deployment. To illustrate this behavioral lack,
a deployment is presented Figure 3
1
based on an
example of Robot application (see Figure 2). Obvi-
ously several concepts instantiations of the considered
RTOS are possible following the application. Accord-
ingly, this illustration is focused on a particular case :
the semaphore sharing.
In this way, the PIM representing the robot ap-
plication combines two periodic activities (a Driver
activity for driving the robot and a Vigil activity for
avoiding barriers). On the one hand, the Driver in-
1
The languages used in Figures 2 to 5 come from the
graphic description of each metamodel (their concrete syn-
tax).
FormalBehavioralModelingofReal-timeOperatingSystems
409
Figure 2: Model of a robot application.
vokes a Moving Strategy program. On the other hand,
the Vigil invokes a Barrier Bypass Strategy program.
Then two wheels allow the robot to move thanks to
two settings updated according to the activities. Fi-
nally each periodic activity owns a priority according
to its role.
The PDM described by RTEPML for this deploy-
ment is a model of the nxtOSEK’s RTOS (nxtOSEK,
2009). nxtOSEK is based on OSEK/VDX standard
(OSEK/VDX Group, 2005) and is widely used in
LEGO MINDSTORMS NXT educational projects.
According to our needs, some resources havebeen de-
fined to be instantiated with regards to the application.
The Task concept is instantiated to represent each ac-
tivity, the Function concept to execute each program
and the Resource (semaphore concept in nxtOSEK)
concept to protect each wheel setting updating.
The PSM results in a model of the application de-
ployed on the nxtOSEK platform. However the con-
current activities which are governed by both appli-
cation and technological concepts do not concretely
appear in the final model. For instance we do not
know whether each Resource (protection for updating
each wheel setting) is available or not (empty or full).
Moreover the state in which each Task (the Driver
T
and the Vigil T) is set cannot be clearly distinguished.
Finally the impact of the properties such as the prior-
ity and the period ofthe tasks is not directly verifiable.
Consequently this illustration shows how difficult
it is to interpret the behavior of the resources instan-
tiated. The behavioral aspect of the generated model
is not observable. In this approach, it is therefore dif-
ficult (if not impossible) to undertake some V&V ac-
tivities to verify the correctness of the deployment.
In order to be able to check the properties on the de-
ployed application, formal behavioral aspects have to
be introduced within the description of the software
platform.
4 STRATEGY OF BEHAVIORAL
FORMALIZATION
In accordance to the previous conclusion, RTEPML
has been enriched to bring a behavioral aspect in the
RTOSs description. On this point, we have extended
the abstract syntax of RTEPML in which executive
concepts have been defined to explicitly describe any
platform.
Once RTEPML extended (newly called
RTEPML
BEHAVIOR), the implementation of
a transformation process could have been put in
place to make the model of the deployed appli-
cation behavioral. To meet the requirements of
V&V activities, the formalization of the PSM is
necessary. RTEPML
BEHAVIOR has therefore been
designed to take into account any formal language.
This solution has been thought in order to make
the transformation independant of other formal
languages.
4.1 Behavioral Consideration
Taking into account the RTOSs behavior necessitates
the identification of the behavior of each resource
(scheduling task, state of a semaphore...). The same
applies to the behavioral description of each service
offered by resources (task terminating, semaphore ac-
quisition...). Lastly, each behavioral impact associ-
ated with each resource property (task priority, task
period, semaphore tokens number...) should also be
treated during the deployment.
To keep on the previous focus, an example of be-
havioral consideration around the semaphore is de-
picted Figure 4. As explained earlier, some concepts
have been added in RTEPML
BEHAVIOR to repre-
sent the behavior of the RTOSs resources. Thus, the
behavioral prototype of the semaphore concept, called
ResourceBehavioralPrototype, has been defined on
the Behavioral PDM. As a reminder, Resource (on
Behavioral PDM) represents the semaphore con-
cept of nxtOSEK RTOS, whereas ”Resource” is the
RTEPML BEHAVIOR metaconcept allowing to de-
scribe any RTOS resource. In other words, Resource
conforms to MutualExclusionResource.
The formalism used to express the behavioral pro-
totype is Time Petri Net (TPN) (see (Boyer and Roux,
2008) for a survey). TPNs have been prefered be-
cause of their expressiveness. This formalism is in-
deed adapted to the description of concurrency activ-
ities such as resources schedulability, resource shar-
ing, call services sequence, time constraints... Fur-
thermore, TPNs proposeknown time extensions (such
as stopwatch TPNs) which allow to model real-time
ICEIS2012-14thInternationalConferenceonEnterpriseInformationSystems
410
Figure 3: Robot application deployment.
schedulings implemented in V&V tools like Romeo
(Gardey et al., 2005) (Lime et al., 2009) or Tina
(Berthomieu et al., 2004).
Figure 4: MER behavior.
As can be seen on Figure 4, a behavioral proto-
type can be composed of several behavior elements.
Indeed, a RTOS resource includes services and prop-
erties. The appearance of roles (defined at meta-
model level) within the Behavioral PDM, enables to
precise certain behavior elements. In the nxtOSEK
semaphore case, serviceBehavioralDescription role
comes to describe the services behavior of getRe-
source and releaseResource, whereas defines role
comes to define the value of the semaphore.
4.2 Deployment with
RTEPML
BEHAVIOR
The behavioral description of RTOSs has been made
possible with RTEPML
BEHAVIOR. Moreover the
possibility to define formally the behavior elements,
contributes to the formalisation of the application de-
ployment. This important point should enable to ver-
ify its correctness before the code generation.
In accordance with the adopted approach, a trans-
formation has been developed to consider explicitly a
behavioral PDM of the targeted RTOS. We have taken
the same robot application from the previous section
to illustrate an example of formalisation of the de-
ployment (see Figure 5). For the sake of simplicity
and clarity, we have reduced the Robot PIM to the
Driver
T instance representing the periodic Driver
activity. Driver
T has an entry point on the Moving
Strategy in which only a service is called for termi-
nating itself.
The involved prototypes in the transformation are
those of the resources for the needs of the application.
Once the prototypes are detected, the latter are cloned
according to each instance of the application. It may
be observed that the cloned prototypes corresponding
to the behavior of the Driver
T and the Moving Strat-
egy (including the service of terminating) have been
composed through this example.
As a result, the TPN Behavioral PSM achieved is
more expressive than the previous one. Its expressive-
ness helps us to easily see its liveness (the different
states of Driver T, its periodicity...).
The composition of the building blocks repre-
sented by cloned behavioral prototypes (semaphore,
task, communication, priority...) make up the gener-
ated model. In TPNs case, the composition is car-
FormalBehavioralModelingofReal-timeOperatingSystems
411
Figure 5: Formalisation of robot application in TPN.
ried out by merging places interfacing all the blocks.
This presents some advantages. The semantics of the
places merging is obvious, while the one of the tran-
sitions merging is ambiguous. Indeed, the fact of as-
signing different time constraints on mergeable tran-
sitions makes their merging difficult. In addition, the
places merging preserves the blocks behavior, which
enables to validate more easily the building.
In this paper, the purpose is neither to describe
exhaustively the modeling of the behavioral proto-
types and the composition of their clones, nor to prove
that it is correct. The aim is only to present the ap-
proach and the methodology which allows it. That is
why no proof of the building validity has been given.
Notwithstanding, this proof is greatly simplified by
our composition approach.
5 BENEFITS OF THE STRATEGY
The main benefits of our methodology are both the
formalization and the moving of the application de-
ployment into the design phase. This should allow to
ensure its correctness before generating code. Bas-
ing our approach on MDE, a formalization of the de-
ployed application model can be carried out. V&V
activities will become applicable thanks to the use of
model-checking tools.
A second benefit is the separation of the skills
highlighted in the second section. The distinction
of modeling artefacts allows both RTOSs behavior
and formal verification experts to interact on the
application deployment (see Figure 6). Thanks to
RTEPML
BEHAVIOR, the specialist of the RTOSs
behavior can be assisted by the one more highly
skilled in the formal verification. The latter one can
also interpret the most adapted formalism following
the case.
Figure 6: Intervention of specialists.
To put our strategy into practice, we have chosen
to verify a necessary and well-known condition in the
RTESs domain: the absence of deadlock. Once the
PSM generated in TPN, the formal model can be pro-
cessed by ROMEO (Gardey et al., 2005). ROMEO
has been developed within the Real-Time team at IR-
CCyN
2
lab. One of the reasons for choosing ROMEO
is that this tool is dedicated to Transition-TPN (T-
TPN) verification, i.e. the time is only assigned to the
transitions. Algorithms were proposed for applying
verification on T-TPN extended to scheduling.
Despite the combinatorial explosion risk of the
generated TPN, ROMEO has given a deadlock out-
come. This diagnosis could seem foreseeable from
the specification since the wheels settings updating is
protected by semaphores. Notwithstanding, this veri-
fication helps us to evaluate the right protocol to use
2
http://romeo.rts-software.org
ICEIS2012-14thInternationalConferenceonEnterpriseInformationSystems
412
for this application. Indeed, the Priority Ceiling Pro-
tocol (PCP) (OSEK/VDX Group, 2005) should have
been chosen to remedy this situation during the de-
ployment.
We have exposed the deadlock case but other be-
havioral properties could have been tested. For in-
stance we could deal with liveness properties such
as termination of actions, occurency of expected
events... But we could also handle time constraints
such as Worst-Case Execution Time (WCET), respect
of a deadline...
6 LIMITS
This experimentation has showed the feasability of
generating a formal model to observe the behavior
of the application deployed on a RTOS. Nonetheless,
the behavior of some of concepts (the notification re-
sources such as the events or the communication re-
sources such as the messages...) are not yet taken into
account within RTEPML
BEHAVIOR. As a result, an
optimization of the latter modeling language must be
widened in order to cover such concepts.
Another limit raised is the interpretation of other
formalisms to model the behavior during the deploy-
ment. Only TPNs have been presented through our
strategy but the implementation of other formal lan-
guages deserves to be examined.
Lastly, a deadlock has been detected during the
deployment. It has been deduced that the deploy-
ment was not correct. That being so, this decision
does not affect the correspondence between the ap-
plication and the targeted RTOS. Indeed, a proto-
col changement (Priority Ceiling Protocol (nxtOSEK,
2009) (OSEK/VDX Group, 2005)) is just sufficient to
prove it. This leads us to say that the framework does
should integrate decision support tools for guiding the
experts involved in the RTOS choice.
7 CONCLUSIONS
Through this study, a methodology has been pre-
sented to consider the behavior of a real-time execu-
tion software platform (RTOS) during an application
deployment. The deployment which enables to im-
plement an application on a specific RTOS, has been
moved at design phase. The purpose of this choice
is to enable the different specialists to intervern more
easily on the deployment following its domain. In-
deed, thanks to the modeling with RTEPML and the
MDE approach, the separation of concerns (applica-
tion and RTOS) has been made explicit.
The description of the RTOSs behavior has
been made feasible by enriching RTEPML
(RTEPML
BEHAVIOR). In addition, each be-
havioral model can be formally described. A
transformation has been developed to generate a
formal model of the deployed application. The
formalization has the advantage of applying V&V
activities for checking the correctness of the deploy-
ment, even before generating the code useful to the
implementation of the application.
However, certain limits of our methodology have
arisen. Future prospects are the subject of our next
works to improveand to check the feasibility of such a
strategy. In accordance to the previous section, the be-
havioral concepts missing in RTEPML
BEHAVIOR
will have to be created. Then, to ensure the genericity
of the formalisation, other languages will have to be
interpreted. Finally, a large number of formal proofs
will have to be written within our framework. This
should contribute to the verification of necessary and
sufficient conditions for validating such deployments.
REFERENCES
Berthomieu, B., Bodeveix, J.-P., Dal Zilio, S., Dissaux, P.,
Filali, M., Gaufillet, P., Heim, S., and Vernadat, F.
(2010). Formal Verification of AADL models with
Fiacre and Tina. In ERTSS 2010 - Embedded Real-
Time Software and Systems, pages 1–9, TOULOUSE
(31000), France. 9 pages DGE Topcased.
Berthomieu, B., Ribet, P.-O., and Vernadat, F. (2004). The
tool tina construction of abstract state spaces for
Petri nets and time Petri nets. International Journal
of Production Research, 42(4).
Boyer, M. and Roux, O. H. (2008). On the compared ex-
pressiveness of arc, place and transition time Petri
nets. Fundamenta Informaticae, 88(3):225–249.
Brun, M. and Delatour, J. (2011). Contribution on the soft-
ware execution platform integration during an appli-
cation deployment process. First Topcased Day.
Davis, J. (2003). GME: the Generic Modeling Environ-
ment. In OOPSLA ’03: Companion of the 18th an-
nual ACM SIGPLAN conference on Object-oriented
programming, systems, languages, and applications,
pages 82–83, New York, NY, USA. ACM.
Gardey, G., Lime, D., Magnin, M., and Roux, O. H. (2005).
Rom´eo: A tool for analyzing time Petri nets. In 17th
International Conference on Computer Aided Verifi-
cation (CAV’05), volume 3576 of Lecture Notes in
Computer Science, pages 418–423, Edinburgh, Scot-
land, UK. Springer.
Lee, E. A. (2003). Overview of the Ptolemy project. Tech-
nical Report UCB/ERL M03/25, EECS Department,
University of California, Berkeley.
Lime, D., Roux, O. H., Seidner, C., and Traonouez, L.-M.
(2009). Romeo: A parametric model-checker for Petri
FormalBehavioralModelingofReal-timeOperatingSystems
413
nets with stopwatches. In Kowalewski, S. and Philip-
pou, A., editors, 15th International Conference on
Tools and Algorithms for the Construction and Anal-
ysis of Systems (TACAS 2009), volume 5505 of Lec-
ture Notes in Computer Science, pages 54–57, York,
United Kingdom. Springer.
nxtOSEK (2009). NXTOSEK Operating System, version
2.10. http://lejos-osek.sourceforge.net/.
Object Management Group (OMG) (2001). Model
Driven Architecture (MDA) Guide, version 1.0.1.
http://www.omg.org/mda/.
Object Management Group (OMG) (2007a). UML Profile
for Modeling and Analysis of Real Time and Embbe-
ded systems (MARTE), second revision submission.
http://www.omg.org/marte/.
Object Management Group (OMG) (2007b). Unified Model-
ing Language (UML) : Superstructure, version 2.1.2.
http://www.omg.org/mda/.
OSEK/VDX Group (2005). OSEK/VDX Operating Sys-
tem Specification, version 2.2.3. http://www.osek-
vdx.org/.
Renault, X., Kordon, F., and Hugues, J. (2009). From aadl
architectural models to petri nets: Checking model vi-
ability. In ISORC, pages 313–320.
Society of Automotive Engineer (SAE) (2004). Architec-
ture Analysis & Design Language (AADL) AS5506,
version 1.0.
Team, M. P. (2004). The metropolis meta model - version
0.4. Technical Report UCB/ERL M04/38, EECS De-
partment, University of California, Berkeley.
The MathWorks (2007). Real-Time Workshop User’s
Guide. The MathWorks Inc., Natick, MA, USA.
Thomas, F., Grard, S., Delatour, J., and Terrier, F. (2008).
Embedded Systems Specification and Design Lan-
guages, Selected Contributions from FDL’07, vol-
ume Embedded Systems Specification and Design
Languages of FDL selected papers, chapter Soft-
ware Real-Time Resource Modeling, pages 169–182.
Springer, Barcelona, Spain, springer science+business
media b.v. edition.
Vergnaud, T. and Zalila, B. (2006). Ocarina, a com-
piler for the AADL. Technical report, Paris, France.
http://ocarina.enst.fr.
ICEIS2012-14thInternationalConferenceonEnterpriseInformationSystems
414