Journal of Systems Architecture 59 (2013) 1013–1014

Contents lists available at ScienceDirect

Journal of Systems Architecture

Guest editorial

Embedded Systems Software Architecture

Embedded Systems are the dominate type of computer systems motive industry, avionics or telecommunication, but also in
today; they span a range from small systems that include a simple new areas such as Internet of Things, e-health, environment,
platform integrated with sensors and actuators, to large distrib- smart houses and smart cites, etc. The Reference Architectures
uted systems consisting of hundreds, or possibly thousand, inten- improve the reusability of methods and artifacts, but also brings
sely interactive nodes. In recent years Software has become the new research challenges such as finding the principles and
most important part of Embedded Systems – it implements the methods for the Reference Architecture deployment with guar-
complex system functionality, is a currier of the system integra- antees for domain-specific functional and extra-functional
tion, and it is enabler of important extra-functional system proper- properties.
ties. In many aspects, software for Embedded Systems has reached  Services Embedded Systems. While software in traditional Embed-
functional complexity of general-purpose software but faces severe ded Systems was encapsulated within a device, and hardly chan-
constraints. Many extra-functional properties and constraints are ged after its deployment, the modern Embedded Systems are
of crucial importance for Embedded Systems: timing characteris- integrated in an extended environment in which external ser-
tics such as response and execution time, performance, manage- vices can be included. A typical example is an infotainment sys-
ment of often limited resources such as energy, memory, tem in cars which provides information about the car itself, but
bandwidth, and dependability properties, such as reliability, safety, also about the traffic, or weather condition, or similar. The
and security. As the complexity of such systems grows, Software extended environment makes the behavior of such systems less
Architecture (SA) becomes means for design and run-time solu- predictable, and that requires new analysis methods that take
tions that ensure the functional and extra-functional properties into account this uncertainty.
will be satisfied. Many established SA methods, theories, lan-  Components in Embedded Systems. A component-based develop-
guages, etc., developed for general-purpose software systems have ment approach is a proven concept for managing complexity,
been successfully deployed in the embedded systems domains, but and has been used also in Embedded System Software for a
the many challenges (e.g. relations between system and software while. Still, many challenges related to predictability and utiliza-
architecture, design and run-time architectures, dynamic architec- tion efficiency vs. reusability are open: For example – can a
tures, reusability of software components, etc.) remain to be reuse of an existing component break the system integrity, or
solved. have impact on timing properties? Since the components, can
be implemented as software executable units, or hardware exe-
1. Challenges cutable units (running of FPGA), there are also questions
whether a software or hardware implementation is better and
Due to increased complexity in software, and a continuous ad- under which condition, or can we have the same component
vances in hardware, new challenges related to complexity manage- model for components implemented as software or hardware
ment and optimal utilization hardware emerge: units?
 Design and run-time Software and System architecture. The mod-
 Modeling and designing extra-functional properties. There exist ern embedded Systems allow dynamic deployment of software,
many theories, methods and technologies for analysis of extra- such as update of the entire or a part of software to get an
functional properties, many of them used in Embedded Systems improved functionality, or to adapt to the environmental con-
domains. These achievements must be now revised and extend text changes. This poses a number of challenges that are not
due to new hardware platforms used in Embedded Systems – solved yet; for example, what is a relation between design-time
for example use of heterogeneous platforms that include multi- and run-time architecture since they can be quite different? This
core CPUs, FPGA and GPUs. The extra-functional properties challenge is also related to dynamic software architectures, e.g.
behave differently on different platforms, so there is a need to applying plug-and-play principles for Embedded Systems by
find methods for specifying, accessing, and measuring these keeping the system integrity, availability and other important
properties. Also, for analyzing them, and for composing them properties.
in the new contexts.  Architecture Description Languages for Embedded Systems. The
 Reference Architectures and their deployment. The importance of new paradigms require new modeling languages, including
reference architectures increases by the increase of different new Architecture Description Languages.
domains in which embedded systems become a dominant part,  Software Architectures for heterogeneous platforms. The new
and in which software becomes the most important component. Embedded Systems architecture becomes heterogeneous. The
This does not happen only in ‘‘classical’’ domains such as auto- heterogeneity is expressed through different executional units

1383-7621/$ - see front matter Ó 2013 Published by Elsevier B.V.

– mutli- and manycore CPUs, FPGAs, GPUs. A software applica- different nodes and, and the system allows a specification of a large
tion can be distributed over these different platforms, and that number of constrains of different types (for example energy con-
may require specific architectural solutions: for example, con- sumption constraints). The article develops a specific algorithm
nectors automatically built, or specific design patterns. Another for fast calculation of an optional solution with a goal to use the
example is related to the software deployment: the same archi- algorithm for a system adaptation during run-time.
tectural model different implementations can be generated, Chih-Sheng Lin, Chao-Sheng Lin, Yu-Shin Lin, Pao-Ann Hsiung, and
depending on the component allocation. Chih-Hsiong Shih: ‘‘Multi-Objective Exploitation of Pipeline Parallelism
using Clustering, Replication and Duplication in Embedded Multi-core
Systems’’ considers the power consumption related to latency and
2. Papers in this special issue throughput. It proposes a heuristic algorithm, called Parallel Pipe-
line Latency Optimization for high performance embedded sys-
This special issue includes six research papers that address tems (PaPiLO), based on clustering, replication and duplication,
some of the mentioned challenges. to minimize latency under power and throughput constraints.
Moazzam Fareed Niazi, Tiberiu Seceleanu and Hannu Tenhunen: ‘‘A Sagar Moreshwar Behere, Martin Törngren, and DeJiu Chen, ‘‘A
Development and Verification Framework for the SegBus Platform’’ reference architecture for cooperative driving’’ addresses a specific
describes development framework for a platform-based design ap- problem in vehicular domain: How to describe and model for
proach, in the context of the SegBus platform. The framework uses collaborative driving (i.e. driving based on information from other
a model-based approach based on UML, which is extended to a Do- vehicular in the traffic, in particular follow a car, or building a
main Specific Language to support the platform modeling. The convoy. The article proposes a reference architecture which
framework provides automated procedures for platform build-up identified the artifacts and the method to architect this type of
and application mapping. applications.
Ashkan Beyranvand Nejad, Anca Molnos, and Kees Goossens: ‘‘A
Unified Execution Model for Multiple Computation Models of Stream- This issue reviewers:
ing Applications on a Composable MPSoC’’ proposes an execution
model that aims at filling the abstraction level gap between the Kenneth Lind, Chalmers Institue of Technology, Sweden
primitives of computation models and the MPSoC (Multiprocessor Miroslaw Staron, University of Gothenburg, Sweden
System-on-Chip) computational models. The execution model is Brahim Hamid, University of Toulouse, France
formalized by defining the execution operations corresponding to Guillermo Rodriguez-Navas, University of the Balearic Islands,
the MoCs’ primitives, proposing a time model of applications exe- Spain
cuting on a composable platform, and presenting different options Patricia López Martínez, University of Cantabria, Spain
of mapping the execution operations to the time model. The mod- Raffaela Mirandol, Politecnico di Milani, Italy
els of computation are mapped to the unified model of execution, Amund Skavhaug, Norwegian University of Science and
and the trade-offs in different mapping options are discussed. Technology, Norway
Franz Dietrich, Jochen Maass, Jens Steiner, Matthias Hagner, Ursula Shuhao Li, Aalborg University, Denmark
Goltz, and Annika Raatz: ‘‘Dynamic Distribution of Robot Control Com- Tiberiu Seceleanu, ABB Corporate Research, Sweden
ponents under Hard Realtime Constraints – Modelling, Experimental Vittorio Cortellessa, University of L’Aquila, Italy
Results and Practical Considerations’’ addresses a problem of distrib- Stefan Stancescu, POLITEHNICA University Romania
uted execution. The application domain is Robotics and the main Lionel Seinturier, University Lille 1, France
concern is related to the real-time requirements. The article pro- Francois Terrier, CEA, France
vides mapping of control tasks to CPU resources while minimizing Severine Sentilles, Mälardalen University, Sweden
bus communication load. In addition to deployment across multi- Aneta Vulgarakis, ABB Corporate Research, Sweden
ple hosts the article introduces aspects of self-management into Antonio Cicchetti, Mälardalen University, Sweden
robot control architectures based on the continuously measure-
ment of worst-case execution times during runtime.
Marc Zeller, and Christian Prehofer: ‘‘Modeling and Efficient Solving
Ivica Crnkovic and Judith Stafford
of Extra-Functional Properties for Adaptation in Networked Embedded
Guest editors
Real-Time Systems’’ addresses a modeling and efficient solving sat-
Embedded Systems Software Architecture
isfaction of extra-functional properties for embedded systems in
automotive systems domains. The components are allocated on