Approaches to Design Complex Software Systems
Chaimae Ouali-Alami
a
, Abdelali El Bdouri, Nisrine Elmarzouki, Ayoub Korchi and Younes
Lakhrissi
b
SIGER Laboratory, Sidi Mohamed Ben Abdellah University, Fez, Morocco
Keywords: Complex software system, Decomposition, Multi-modelling, Design.
Abstract: Despite the evolution of design techniques in the field of software engineering, building complex computer
systems remains a very difficult task for the modelling, design and analysis team due to the complexity,
richness of information and greatest scope. The modelling of complex software systems is a very rich and
immense field of research. Decomposition of complex systems into smaller pieces or components is the best
available tool to reduce complexity and dimension also to facilitate construction by humans. System
decomposition is a fundamental component of the systems modelling process. In this paper, we will present
some methods and techniques of model decomposition. We will explain the importance of decomposition,
itemize decomposition methods as we will focus on a set of approaches.
1 INTRODUCTION
The ubiquity of technology in all fields; computer
science, mechanics, industry, economics, and
commerce, requires a lot of effort in terms of analysis,
modelling, and design. Conceptual models have been
created: (1) to understand a problem and its context,
(2) to provide a basis for analysis and development.
This task facilitates the choice of the most suitable
solution. With the increased need of users and more
deep requirements, the complexity of the studied
problem increases more and more. Solve such
complex and difficult systems become an urgent
need. The system decomposition method is the best
way to manage and analyse a complex system by
decomposing it into elementary problems to pick out
an adequate solution.
To attack a complex system, it is necessary to
determine the needs of the actors although the
technical requirements (Lakhrissi, 2010) whose
construction of the global model remains a difficult
task, despite the evolution of analysis and design
techniques in the field of software engineering. Multi-
modelling approaches are model-oriented
approaches, using model development separately. To
discuss these so-called model-oriented approaches, it
a
https://orcid.org/0000-0003-3862-9149
b
https://orcid.org/0000-0003-2718-7090
is important to return to object-oriented approaches
(Abiteboul, 1991) (Kriouile, 1995) (Harrison, 1993),
that is limited when faced with a multidimensional,
complex, and highly parallel requirements.
To cope with this complexity, we use, more and
more, so-called multi-model modelling approaches.
This method offers good practices of decomposition.
This paper focuses exclusively on the design of
the complex system. In the second section, we present
the decomposition of systems and some good
practices. Then in the third section, we present four
multi-modelling approaches: (1) point of view
modelling, (2) role modelling, (3) aspect modelling,
(4) subject modelling. Finally, the last section
presents the conclusion and possible future works.
2 DECOMPOSITION METHODS
When we are facing a complex system, we need to
break them down into sub-systems. For this, we must:
(1) estimate the complexity and the size of the sub-
systems, (2) scale the system into pieces, main
functions or main sub-systems, (3) split each system
into sub-functions or components.
Ouali-Alami, C., El Bdouri, A., Elmarzouki, N., Korchi, A. and Lakhrissi, Y.
Approaches to Design Complex Software Systems.
DOI: 10.5220/0010740200003101
In Proceedings of the 2nd International Conference on Big Data, Modelling and Machine Learning (BML 2021), pages 533-537
ISBN: 978-989-758-559-3
Copyright
c
 2022 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
533

In his paper ( Chiriac, 2011) Chiriac proposes three
approaches. The first one refers to the
Disassembly/Assembly method following a set of
instructions: (1) visually inspect the system, (2)
identify sub-assemblies that can be easily removed,
(3) remove sub-assemblies that are probably
represented by large pieces, (4) record them in a DSM
(Design Structure Matrix). The second proposed
approach is functional decomposition that
implements a decision by disciplines up to
components to fill a DSM. The last approach is a
service-based decomposition.
The Design Structure Matrix (DSM) is a compact
representation of a system or project. It consists of a
list of subsystems, activities and their associated
information exchange patterns. The information
pieces that are required to start an activity are referred
to as parameterizations. The output information is
also used by other tasks within the matrix. There are
three types of DSM configuration:
 Parallel: fully independent design elements,
 Sequential: the second parameter depends on
the output of the first,
 Coupled: all parameters have a dependency
between them.
The figure 1 bellow illustrates those types with
their graph representation and DSM representation
(
Guenov, 2004).
Stephen proposes a modelling process (
Topper,
2013) through the creation of a set of artefacts:
 Description of the system and its environment
"Domain Model".
 Preparation of a diagram of "Use Case".
 Division of the use cases in detail by showing
the flow of activities and transitions state
between components.
 Broaden the domain model by adding attributes
and operations. It is an iterative process that
tires to create the final result and to complete
the solution within its problem domain.
Figure 1: DMS configurations
The notion of decomposition presented in
(
Guenov, 2004) has two objectives: (1) basics for the
complex systems analysis, (2) basics for the complex
systems design. According to Yair Wand, he quoted
in his paper (
Wand, 1990) a very good point in the
principles of good decomposition concerning the
subsystems «weakly coupled». Before the definition
of the sub-systems developed by the designers, he
defined a set of transformation rules before the
decomposition to check the complete system design
afterwards.
Good design practice is axiomatic design that
insists on (1) design decision in domain and hierarchy,
(2) maintenance of functional requirements, (3)
minimization of design information content.
Quite a lot of research supports multi-agent
systems (MAS), MAS is recognized as abstraction
technologies for the modelling and construction of
complex, and autonomous systems called useful and
efficient. This poses a challenge due to the
distribution and openness of these systems, the
autonomy of ownership, as well as the complex
nature of the internal functionalities of agents and
interactions between agents(
Alaca, 2021) .
To develop MAS using model-led engineering
(MDE) techniques, the most popular way is Domain-
Specific Modelling Language (DSML) (
Sredojević,
2018) with an Integrated Development Environment
(IDE).
3 MULTI- MODELLING
APPROACHES
When tackling the complexity of huge software
systems, separation of concerns is important for
keeping the event process, the produced models and
therefore the code manageable. The separation of
concerns are often wiped out alternative ways, but the
objective is always the same: having the ability to
spot relatively independent “parts”.
In this section, we present the four major multi-
modelling approaches: Views modelling, Aspects
modelling, Subject modelling and Role modelling.
3.1 Point of View Modelling
VBOOM (View-based Object-Oriented Method)
(Kriouile, 1995) is an object-oriented programming
method that use the point of view concept and the
view-oriented approaches based on UML(
Anwar,
2009)
.
BML 2021 - INTERNATIONAL CONFERENCE ON BIG DATA, MODELLING AND MACHINE LEARNING (BML’21)
534

The VUML language (View-based Unified
Modelling Language) (
Nassar, 2005) is a UML profile
based on the point-of-view modelling approach to
deal with the limitations of object-oriented
programming. Primarily the implementation of views
by multiple instances. This approach is based on the
concept of point of view to analyse and design a
software system, in particular, the establishment of
several partial models. It is applied especially to an
information system which is often characterized by a
strong interaction with users; whose actor is defined
(Lakhrissi, 2010) as a human being or an entity that
interacts with the system. A point of view can be
defined as a vision of an actor on a part of the whole
of the system. A View is a modelling entity on which
the point of view of the actor is applied. Generally,
the patterns associated with software developments
could be several types of actors:
 development actors when designing and
implementing the pattern :Analyst, designer,
programmer, tester, and maintainer.
 final-users when instantiating the pattern in a
given application domain (figure 2).
Figure 2: Hierarchy of viewpoints
In view approaches, the decomposition level is
different from that adopted by the aspect approaches.
This is often decomposition consistent with the views
of actors of the system. The views are developed
independently of any other sub-system and without
making any distinction between the essential
functionalities and the cross-functionalities. The
results of this process of decomposition may be a set
of entities described by the subjective views of the
actors of the system.
3.2 Aspect Modelling
After an advanced search for aspect-oriented
programming, this approach has reached maturity
(
Chavez, 2002) . There is a need to address a fair
amount of interdependent design problems to have
software. The purpose of modelling by aspect (
France,
2004)
is to create software that takes into account
safety, fault tolerance (
Sutton, 1999) and to treat
reliability problems. In a particular way, this problem
can make difficult or impossible to solve other
problems. That pushes the developers to adopt aspect-
oriented modelling (AOM) approach that facilitates
the task of exploring other ways to address concerns
during software modelling. Then this approach
focused mainly on the balance between the reliability
of the keys and other concerns during the early stages
of development. This can help developers better
manage product risks through the identification and
early resolution of conflicts and behaviours that
address different concerns. In this context, the authors
define a Concern as a problem associated with the
desired objective. The Concern Solution Model is a
model that describes how the concern is addressed
(
Chavez, 2002) ( France, 2004) .
We distinguish between two types of concerns:
 A concrete concern: includes solutions that can
be expressed in functional and structural terms
in a model (Access control and error recovery).
 A qualitative concern: based on the qualities or
attributes of a system (related to system
performance and memory usage).
In general, traditional design has difficulties in
creating and above all in having complex and reliable
systems evaluated due to several factors such as the
difficulty of changes related to these concerns, the
dissemination of information has become more
difficult, especially with the multiplicity of design
teams (
Noda, 1999) . The mechanisms provided by
multidimensional concern separation techniques
(MDSoC: Multi-Dimensional Separation of Concern)
can reduce the cognitive burden of creating and
evolving reliable software (
Tarr, 1999) .
The aspect method decomposes the system into
functional units and non-functional units (
Andersen
1992)
.
3.3 Role Modelling
In object-oriented programming, a class is a type of
data that is complemented by a formal or informal
OCL description of their behavioural effect, whereas
when faced with a method that is part of an activity
involving several communicating objects. It is
difficult to understand this description and difficult
too to add a description for each relevant class. That
lead to the emergence of role-based modelling, it
allows the designer to focus on several aspects or a
single aspect with several independent levels of
detail. The designer has to build a role model for each
Approaches to Design Complex Software Systems
535

activity or task or to build several role models for the
same activity at different levels of detail. This shows
that role modelling is more adjacent to the functional
approach than the objects approach.
A role model can be defined as a design unit that
identifies a structure that includes two or more
entities in external role interaction. The latter is
considered as a requirement or liability (
Kendall, 1999)
(Kristensen, 1996). Saying that, an object can be
considered as a particular role, which explains the
definition of role as a temporary point of view (
Riehle,
2000). Moreover it is a type that describes the view
that an object vis-à-vis another which allows an
object to play different roles at a given moment
(
Ossher, 1995).
In the role approach, it is to represent an entity of
the model through multiple objects. Each object
models an exact role played by an entity. Unlike
modelling by views, roles are objects resulting from
local entities subjective views without being linked to
actors of the system. The various subjective views of
the system represent all of the appliance concerns,
and every concern is represented by the various roles
and their interactions.
3.4 Subject Modelling
Harrison introduced another technique of separation
of concern (
Harrison, 1993): Subject-based
programming based on multidimensional separation
of concerns. It is defined as a visibility of the world as
a whole by a particular application or tool, and so a
subject and a generalized perception, whose
characteristics of an object can be applied to any other
object. An object can activate several subjects and a
subject can be activated for several objects. This
approach present a group of advantages like
unscheduled extension and composition,
decentralized class development, also a code that
implements a feature that will be programmed as a
subject [23].
The subject approach extended by the MDSoC
approach (Multidimensional Separation of Concerns)
offers a decomposition of the system into more
arbitrary dimensions, where each dimension may be a
collection of particular concerns. We mention concern
within the broad sense, without differentiating
between basic concerns or crosscutting concerns.
4 SYNTHESIS
The concept of decomposition methodology is to
interrupt a posh classification task into simpler and
more manageable steps, which can be solved by using
existing induction methods.
The main interest of the modular approach is to
facilitate the construction of systems by allowing 1)
to write the module with little knowledge about the
other code modules, and 2) to replace one or more
modules without reassembling all the system.
The construction of the system is more
understandable, manageable, and maintainable. One
limitation of the approach is we do not have the
possibility to personalize a module, it means to allow
make many variations of an existing module.
The Architectural Description approach focuses
on designing systems by assembling architectural
bricks. Three concepts of architectural bricks are
popularly accepted components, connectors and
configurations. Architectural bricks are modelling
portions of an abstract system without defining their
implementations. So we can use these portions when
the system is in the analysis phase to allow create the
architecture description. This description can help us
later to make simulations on the system or to allow
other people to analyse the system. Therefore, the
variety of components created compatibility
problems, portability, multiplicity, complexity and
lack of standardization.
5 CONCLUSION
This paper has described a set of approaches to
support system analysis and system modelling and
system design using the decomposition process.
Current approaches provide good support for
modularizing systems along a few dimensions. They
can significantly enhance support for separation of
concerns, and also can help developers to manage the
complexity of building complex software. The
process of decomposition involves a step-by-step
process where a global representation of a system is
proposed (Models Composition). This step is
necessary to get a complete global representation of
the system under construction.
The concept of the composition (
Elmarzouki, 2016)
model is a challenging topic to discuss as it concerns
to the definition of new approaches.
REFERENCES
Lakhrissi, Y., 2010. Intégration de la modélisation
comportementale dans la conception par points de vue
(Doctoral dissertation, Universite Toulouse le Mirail-
Toulouse II).
BML 2021 - INTERNATIONAL CONFERENCE ON BIG DATA, MODELLING AND MACHINE LEARNING (BML’21)
536

Abiteboul, S. and Bonner, A., 1991. Objects and views.
ACM SIGMOD Record, 20(2), pp.238-247.
Kriouile, A., 1995. VBOOM, une méthode orientée objet
d’analyse et de conception par points de vue.
Unpublished doctoral dissertation, University
Mohammed V, Rabat, Morocco.
Harrison, W. and Ossher, H., 1993, October. Subject-
oriented programming: a critique of pure objects. In
Proceedings of the eighth annual conference on Object-
oriented programming systems, languages, and
applications (pp. 411-428).
Chiriac, N., Hölttä-Otto, K., Lysy, D. and Suh, E.S., 2011.
Three approaches to complex system decomposition. In
DSM 2011: Proceedings of the 13th International DSM
Conference.
Guenov, M.D. and Barker, S., 2004, January.
Requirements-driven design decomposition: A method
for exploring complex system architecture. In
International Design Engineering Technical
Conferences and Computers and Information in
Engineering Conference (Vol. 46962, pp. 145-151).
Topper, J.S. and Horner, N.C., 2013. Model-based systems
engineering in support of complex systems
development. Johns Hopkins APL technical digest,
32(1).
Wand, Y. and Weber, R., 1990. An ontological model of an
information system. IEEE transactions on software
engineering, 16(11), pp.1282-1292.
Alaca, O.F., Tezel, B.T., Challenger, M., Goulão, M.,
Amaral, V. and Kardas, G., 2021. AgentDSM-Eval: A
framework for the evaluation of domain-specific
modeling languages for multi-agent systems. Computer
Standards & Interfaces, 76, p.103513.
Sredojević, D., Vidaković, M. and Ivanović, M., 2018.
ALAS: agent-oriented domain-specific language for the
development of intelligent distributed non-axiomatic
reasoning agents. Enterprise Information Systems,
12(8-9), pp.1058-1082.
Anwar, A., 2009. Formalisation par une approche IDM de
la composition de modèles dans le profil VUML
(Doctoral dissertation, Thèse de doctorat, Université de
Toulouse).
Nassar, M., 2005. Analyse/conception par points de vue: le
profil VUML (Doctoral dissertation).
Chavez, C. and Lucena, C., 2002, April. A metamodel for
aspect-oriented modeling. In Workshop on Aspect-
Oriented Modeling with UML (AOSD-2002).
France, R., Ray, I., Georg, G. and Ghosh, S., 2004. Aspect-
oriented approach to early design modelling. IEE
Proceedings-Software, 151(4), pp.173-185.
Sutton Jr, S.M., 1999. Stanley, J., & Sutton, M. (1999,
November). Multiple Dimensions of Concern in
Software Testing. In First Workshop on Multi-
Dimensional Separation of Concerns in Object-oriented
Systems (OOPSLA’99).
Noda, N. and Kishi, T., 1999, November. On Aspect-
Oriented Design-Applying Multi-Dimensional
Separation of Concerns on Designing Quality
Attributes. In First Workshop on Multi-Dimensional
Separation of Concerns in Object-oriented Systems
(OOPSLA’99).
Tarr, P., Ossher, H., Harrison, W. and Sutton, S.M., 1999,
May. N degrees of separation: Multi-dimensional
separation of concerns. In Proceedings of the
International Conference on Software Engineering
(IEEE Cat. No. 99CB37002) (pp. 107-119). IEEE.
Andersen, E.P. and Reenskaug, T., 1992, June. System
design by composing structures of interacting objects.
In European Conference on Object-Oriented
Programming (pp. 133-152). Springer, Berlin,
Heidelberg.
Kendall, E.A., 1999, October. Role modelling for agent
system analysis, design, and implementation. In
Proceedings. First and Third International Symposium
on Agent Systems Applications, and Mobile Agents
(pp. 204-218). IEEE.
Kristensen, B.B., 1996. Object-oriented modeling with
roles. In OOIS’95 (pp. 57-71). Springer, London.
Riehle, D., 2000. Framework design: A role modeling
approach (Doctoral dissertation, ETH Zurich).
Ossher, H., Kaplan, M., Harrison, W., Katz, A. and
Kruskal, V., 1995, October. Subject-oriented
composition rules. In Proceedings of the tenth annual
conference on Object-oriented programming systems,
languages, and applications (pp. 235-250).
Elmarzouki, N., Lakhrissi, Y. and Elmohajir, M., 2016,
March. A study of behavioral and structural
composition methods and techniques. In 2016
International Conference on Information Technology
for Organizations Development (IT4OD) (pp. 1-6).
IEEE.
Approaches to Design Complex Software Systems
537