Research and Integration Activities for the "Adaptive Real Time" cluster

Flexible Resource Management for Real-time Systems
JPRA-Cluster Integration

Resource management in many embedded systems application domains (including consumer electronics, robotics, automotive and telecommunications) faces the challenges of highly variable resource demand, highly variable resource availability, distributed architecture, and strong requirements for resource.

Hence, the objective of the research is to exploit the excellence of the different teams for advancing the state of the art in the theory and practice of techniques leading to adding the required flexibility to the resource management in these systems. The challenge is to develop an efficient resource manager that can be adopted in next generation kernels to perform adaptive QoS control of time sensitive applications with dynamic characteristics.

The former activities “Flexible Scheduling Technologies” and “Adaptive Resource Management for Consumer Electronics” have been merged into this activity.


In some application domains, such as multimedia, applications are very expensive in terms of resource consumption. In other applications domains, such as automotive, mobile telephony or even building automation, the resources are scarce and there is a growing pressure to integrate resources even further and optimize their use. In both cases, timeliness directly relates to user perceived quality, e.g., smoothness of the video stream. Furthermore, efficient resource usage is key issue not only for cost considerations, but also for competition on a feature bases: better resource usage – more features.

Both resource demands, e.g., MPEG-2 video streams, and resource availability, e.g., available bandwidth on wireless links, fluctuate rapidly and unpredictably; worst case assumptions will lead to extreme over provisioning. Consequently, methods for adaptive resource management are required.
Trading resource usage (processing, communication and memory/storage, inter-device and intra-device) against offered output is known as QoS (Quality of Service). The different resources cannot be considered separately, interferences and inter-resource tradeoffs have to be taken into account because they affect the application output. The tradeoffs have to be made at different time scales, in order to match the time scales of the system dynamics.

Theory for independent scheduling algorithms is well defined in the areas of event triggered and time triggered systems, but few theoretical results have been achieved in trying to integrate these approaches. Some partial results exist for simplified architectures, but it is necessary to enhance them by taking into account all of the requirements of modern real-time systems including distributed ones. In addition to the development of theory, a framework needs to be built in order to allow a flexible way to handle different scheduling algorithms for diferrent kinds of resources, and evaluate their applicability to real application domains.

Previous Work

IIn the first period, the technical results were achieved in the following areas: video stream demand analysis, identification of scheduling algorithms and kernel mechanisms for stream adaptations based on integrated, flexible scheduling; adaptive resource management for network bandwidth management, multi resource management, in particular with respect to cache aware scheduling; middleware support for QoS management.

Furthermore, the ART cluster has been in active contacts with relevant industry to gather understanding of realistic requirements and to identify research topics and baselines relevant for industrial and academic research. Partners has been giving presentations at the Philips Software Conference – Real-time Workshop and had meetings with Nokia, Ericsson mobile platforms and Visual tools from Spain. The goal has been to go as far as possible towards the actual engineers for better understanding and prepare for a specific industry – academia workshop with selected participants.

New scheduling mechanisms for integrating overload management techniques with energy-aware strategies were investigated in the context of real-time systems. The new scheduling mechanisms were analysed to guarantee timing constraints while minimizing energy consumption, and a kernel infrastructure was developed into the Shark operating system in order to facilitate their implementation. Moreover, the assessment of the (m,k)-firm model was done for its implementation over the FTT-Ethernet protocol.

Problems Tackled in Year 3

The objectives of the activity in the second year crystallised around the following issues:
  • Temporal Constraints for Video streaming. Philips and TUKL have studied temporal constraints of video streaming. As sources for the constraints we looked into semantic stream dependencies from MPEG decoding, as well as the temporal impact of devices and their resources in the end-to-end delivery chain of a stream. The work was carried out with industrial partners in the area. The results will feed into other activities in the cluster, in particular w.r.t. to scheduling and networking.

  • Integrated real-time scheduling and cache management. Philips and TUKL have continued work on integrating real-time scheduling and cache management on multiprocessor platforms. To this end, we have carried out experiments to study cache behaviour on the actual platform and formulated a number of scenarios with increasing complexity. A joint PhD student is carrying out the work.

  • Adaptive service configuration for Quality-of-Service aware middleware . In order to support dynamic services with adaptive QoS requirements, we have proposed a dynamic scheduler which is able to react to load variations. Isolation between different services is still achieved by guaranteeing a minimal service quality to accepted services and by an efficient overload control that considers the challenges and opportunities of dynamic distributed embedded systems. This scheduler has also been extended taking into consideration simple dependencies between services’ QoS attributes.

  • Server Based Flexible Scheduling.
    Schedulability analysis techniques were developed for server-based systems that can be used to schedule different kinds of flexible timing requirements, such as those needed to integrate control systems with multimedia activities. In particular, this work focused on hierarchical scheduling analysis and design techniques A further issue was the dimensioning of the parameters of a server for minimizing the average response time of the served activities. A statistical approach was addressed in order to compute the probability of missing a given deadline. Partners were SSSA, Cantabria, TUKL of the cluster and the partners of the FIRST and FRESCOR EU STREP consortia.

  • Adaptive resource management for networks.
    Work concerned the analysis of the achievable QoS guarantees in wireless networks. After defining a proper model for the main resources (i.e., CPUs, disk and network), a number of existing scheduling algorithms for the three types of resources were analysed under fluctuating workload to evaluate their behaviour in terms of service guarantee. Then, the achievable end-to-end QoS guarantees were investigated as a function of the guarantees provided by the underlying resource schedulers. Further activities dealt with network protocols to efficiently support dynamic bandwidth management with strict QoS guarantees in Ethernet- based networks, which is still an important networking technology in the field of distributed multimedia systems. A wireless time-token communication protocol that allows providing real- time guarantees for real-time messages and tune the allocated bandwidth according to the required QoS was developed. Aveiro, Porto, SSSA, and TUKL carried out work.

  • Adaptive service configuration for Quality-of-Service aware middleware.
    Complex dynamic real-time scenarios may prevent the possibility of computing optimal service configurations before execution; an iterative refinement approach with the ability to trade off deliberation time for the quality of the solution was specified. The approach is to quickly find good initial solution and to propose heuristic evaluation functions that optimize the rate at which the quality of the current solution improves as the algorithms have more time to run.
    The work also addressed the problem of dynamically changing system conditions, allowing the system to make QoS adaptation decisions in response to fluctuations in the nodes service load, under the control of the user. Monitoring the stability period and resource load variation of Service Level Agreements for different types of services was used to dynamically adapt future stability periods, according to a feedback control scheme. Work was done by Madrid and Porto.

  • Middleware.
    System adaptation requires full knowledge of the system state, therefore work was also carried out in a framework to gather actual resource usage information, and interact with the operating system, extending the traditional POSIX trace model with a partial reflective model for operating system monitoring. The work was done by Porto.
    HOLA-QoS is a framework for managing QoS and resources and it has been used in media processing which UPM and UC3m have developed jointly. It is implemented as a layered architecture, so that layers can be replaced, as far as the API is kept. The higher layers are meant to deal with quality, while lower layers are mainly related with resource management. In particular, the lowest layer is intended to manage budgets or resource shares assigned to applications. This layer has provided accounting and enforcing facilities to ensure that budgets are guaranteed. Sometimes this functionality is provided by what is called resource kernels.
    Cluster partners have developed kernels that provide these facilities and, hence, could be suitable to act as the lower layer of a HOLA-QoS based system. Work that it was under development was to port HOLA-QoS on top MARTE (Cantabria) and SHARK (Pisa) kernels. One result of this work is the possibility of experimenting with the adaptation techniques that these advanced resource kernels provide.

  • Resource availability prediction.
    The resources typically used in-home entertainment applications (e.g., video/audio streaming) exhibit fluctuating availability. It is desirable to have mechanisms for indicating the available bandwidth during system runtime.
    A comparative analysis of bandwidth estimation techniques for WiFi links has been carried out. In particular, the analyzed estimation techniques include several statistical and control-based algorithms. The analysis has identified the best suitable techniques taking into account the specific behavior of WiFi links. Work was carried out by UPC and TUKL

  • Requirements for integrated-resource scheduling framework.
    A workshop on “Requirements for Flexible Scheduling in Complex Embedded Systems” was held in Massy (Paris) in June 2006, with the objective of developing a set of requirements for building a flexible scheduling framework for applications demanding various types of tasks, constraints, and scheduling paradigms within the same system, and paying attention to the integration of multiple resources. The workshop was very sucessful and brought together 20 participants.

  • Baseline for integrated-resource scheduling framework.
    The FIRST (Flexible Integrated Real-Time Scheduling Technologies) IST project that finished in 2005 produced as its main result a contract-based scheduling framework, called FSF that was capable of scheduling multiple application components with various kinds of requirements for CPUs and, to a limited extent, for networks in distributed systems. FSF defines the basis of a new flexible scheduling strategy, and opens a new line of future research in design methodologies that help in translating application requirements into the best possible set of contracts to achieve the highest level of resource utilization and quality. This framework was selected as the baseline for the more ambitious framework that is being developed in this activity and that will take into account the integrated scheduling of multiple resources. The new framework will also include enhanced on-line schedulability analysis and derivation methods for server parameters, as well as the development of high-level transaction management middleware. It will be fully implemented on top of several networks and network protocols, to make it usable for distributed applications (
    The FTT framework, in which a master manages the synchronous activities in a distributed system or cluster, was also extended to micro-segmented switched Ethernet-based distributed systems, having revealed potential to provide efficient support to the contract model, to dynamic QoS management and to integrated resource scheduling in distributed environments. The VTPE protocol (Virtual Token-Passing Ethernet), which supports event-triggered communication with real-time guarantees and high bandwidth utilization, was also extended with appropriate mechanisms to support isochronous traffic, more adequate for some applications, e.g. multimedia transmission. Moreover, the impact of flexible scheduling on dependability for distributed safety-critical applications was also assessed using FTT-CAN and appropriate mechanisms were developed, namely master replication, fail-silence enforcement and a protocol to minimize inconsistencies during updates of the masters table.
    Work was also carried out in a framework that facilitates the distribution of resource intensive services across a community of nodes, forming temporary coalitions for a cooperative execution. In this framework, the system must able to react to load variations, degrading its performance in a controlled fashion if needed. Server-based scheduling approaches were developed to provide isolation between different services and an efficient overload control, considering the resource requirements of the supported services.

  • New theoretical developments.
    The contract-based scheduling framework needs to be implemented using a specific scheduling strategy, and the most effective approach for this case is the server-based hierarchical scheduling in which an application or application component is scheduled over a protected bandwidth-preserving server (such as a periodic server, a sporadic server, or a constant bandwidth server) and individual threads in that component are scheduled by a higher-level scheduler that uses the bandwidth provided by the server. Theory was developed towards being able to analize such scheduling schemes.
    Work was done by the University of York together with Technische Universiteit Eindhoven (TU/e) on the analysis underpinning the use of CAN in real-time systems.
    SSSA developed the following theoretical results: energy-aware scheduling algorithms for processors with dynamic voltage scaling and discrete frequency levels; a method for minimizing the deadline of periodic tasks with the objective of reducing delay and jitter; a general methodology for performing sensitivity analysis of fixed priority periodic systems with configurable periods and computation times, allowing the system designer to derive the feasibility region of a task set and compute the maximum parameter variations that keep the system feasible
    Work carried out at the University of Aveiro also exposed a couple of anomalies related to the definition of critical instant in hierarchical scheduling scopes found in communication systems that led to optimistic worst-case response time analysis in the past. Adequate methods were devised to cope with such anomalies.
    The Polytechnic Institute of Porto provided new theoretical developments on: a new multiprocessor scheduling approach with a higher utilization bound and with few preemptions, able to trade the utilization bound for preemptions; new flexible admission control algorithms for IEEE 802.15.4 networks improving the bandwidth utilization compared to the explicit allocation used in the IEEE 802.15.4 protocol; a new server-based scheduling approach for handling isolation and overload control on distributed cooperative systems.

  • Flexible architectures and communication protocols for networks used in distributed embedded real-time systems.
    This research was partly done in collaboration between the following ARTIST 2 partners: University of Pavia, University of Catania (affiliated partner) and Malardalen University, Sweden. It consists of the following activities:
    • Integration of networked subsystems in a resource constrained environment.
    • Facilitating subsystem integration by decoupling priority and identifier in CAN messages.
    • Interconnection of real-time networks in factory automation and in the automotive domain.
    • Design issues and transmission protocols for wireless networks used in factory communication
    • Bluetooth to support real-time traffic in factory communication.
    • Modelling of wireless real-time communications for land monitoring systems.
    In addition, a joint work was done by Pavia, Pisa and Aveiro to support connectivity tracking in mobile ad-hoc wireless networks subject to real-time constraints.

Current Results

Application requirements (all)
We have expanded the collection of application requirements beyond the video streaming domain. Activities were carried out in particular together with the FRESCOR project, including a joint requirements workshop and a meeting with the industrial advisory board of FRESCOR on the topic. A previously endangered workshop between scientists and engineer in the domain of media processing has been revived and will be held Nov 15,16 2007.

Integrated CPU scheduling and cache management (TUKL, NXP (formerly Philips)
NXP and TUKL have continued work on integrating real-time scheduling and cache management on multiprocessor platforms. We have carried out thourough experiments to study the impact of cache usage and scheduling on predictability on the actual platform. We developed a first algorithm with exhibits encouraging behavior.
A joint PhD student is carrying out the work. Publications are in preparation and have been submitted.

Architectural model of the flexible scheduling framework (All)
An architectural model of a flexible scheduling framework has been developed. The framework is capable of handling multiple concurrent activities with different criticality and timing in the same system, integrating the management of different kinds of resources such as processors, networks, memory, energy, and shared objects with time protection. This framework provides the ability to compose several applications or components into the system, and to flexibly schedule the available resources while guaranteeing hard real-time requirements. The framework is independent of the underlying implementation, and can run on different underlying scheduling strategies. It is based on establishing service contracts that represent the complex and flexible requirements of the applications, and which are managed by the underlying system to provide the required level of service. The framework provides mechanisms to distribute the share of resources among those application parts that are capable of using them. The distribution is made using QoS parameters specified by the application. The framework also separates the management of resources from the actual scheduled elements, and this independence eases the dynamic allocation of resources, which is specially useful for integrating component-based design methods into the framework. An interface between this framework and the applications that run on top of it has been designed.

The work has been carried out by the members of the FRESCOR EU project, together with other members of the ART cluster, in particular UPM and Aveiro.. Of course it is necessary to distinguish those resources that we can manage for scheduling purpose from those that we can just model, or that do not require specific scheduling (

Implementation of the framework (UC, UPVLC, CTU, SSSA, Aveiro)
The flexible scheduling framework described in the previous clause has been implemented for different execution platforms and networks. UPVLC and Cantabria have implemented the framework on RT-Linux and MaRTE OS. SSSA has implemented it on top of Linux and the AQUOSA resource management middleware. An adaptation layer has been created to ease the porting to fifferent platforms. Aveiro has implemented the framework on the FTT-SE protocol [12] [42] [43. Cantabria and UPVLC have implemented it on top of Ethernet networks, with two different kinds of protocols: RT-EP and a time-division multiplexing protocol. CTU is implementing the protocol for wireless networks. York has implemented an on-line schedulability analysis module for the framework, together with an API that makes it possible to plug in other schedulability analysis techniques. The separate implementation on processors and networks requires integration that will be addressed in the following year. An adaptation layer for network protocols has been defined for this purpose. A component-based framework is being integrated together with the flexible scheduling framework [5][6], and schedulability analysis tools [44]. Information on the framework may be found at:

New theoretical developments (York, SSSA, Aveiro)
A number of theoretical developments have occurred during this last year.
• For hierarchical scheduling, a resource sharing model has been defined and the scheduling equations appropriate for this model have been derived and verified. This is applicable to fixed priority scheduling and builds upon one of last year’s developments.
• For systems that allow online/dynamic admissions, efficient scheduling algorithms for fixed priority scheduling have been developed and verified during experimentation. These algorithms allow spare capacity to be allocated to applications in an effective and efficient way, thereby enhancing the adaptability of such applications
• For hierarchical systems that are implemented on an EDF scheduler, necessary and sufficient analysis has been developed that mirrors the fixed priority approach, and thereby allow mixed EDF and fixed pri systems to be deployed.
• For the CAN communication protocol, a previously unknown error in the standard way industry verifies CAN-based systems has been identified and revisited analysis has been developed and published.
• Jitter-aware scheduling algorithms have been proposed and compared at SSSA. The main contribution in this area was to develop two algorithms, with different performace and complexity, to reduce finishing time jitter in control tasks. The proposed techniques exploit deadline reduction mehtods to limit the active execution window of jitter sensitive tasks. Schedulability analysis has been derived and performance has been evaluated with respect to other existing techniques. A more theoretical work has been carried out at SSSA to identify the space of feasible deadlines in a set of periodic real-time tasks.
• Aveiro continued the work about anomalies related to the definition of critical instant in hierarchical scheduling scopes found in communication systems [17].
• Dynamic QoS management requires a fast schedulability analysis to be executed on-line. Pratical analysis are those based on utilization tests but these are rather inefficient. A new bound, called local utilization bound, was developed in collaboration between Aveiro and the University of Pennsylvania, which improves substantially the efficiency of such bounds when the reconfigurations space is limited [11].

Design of a quality of service manager (SSSA, TUKL, UPM, Cantabria, Aveiro, UPVLC)
HOLA-QoS is a framework for managing QoS and resources and it has been used in media processing which UPM and UC3m have developed jointly. It is implemented as a layered architecture, so that layers can be replaced, as far as the API is kept. The work that has been finished is to replaced the lower levels of HOLA-QoS with two kernels with resource management facilities (also called resouce kernels): MARTE (Cantabria) and SHARK (Pisa). The integration has been validated with some test applications. The final part of this work is the integation of the higher levels of HOLA-QoS that is currently being redisengeed. One result of this work is the possibility of experimenting with the adaptation techniques that these advanced resource kernels provide. Some publications on HOLA-QoS can be found at
The AQuoSA framework has been adapted for use with the FRESCOR framework. ,
UPVLC (group from Alcoy) and Aveiro developed a dynamic QoS manager for distributed multimedia systems (MJPEG streams) [47]. Moreover, a method was developed to adapt VBR streams (as generated by the MJPEG sources) to CBR channels (as provided by FTT-SE) [45].

Architecture for dynamic service composition (UC3M, Aveiro)
The group at UC3M (affiliated to UPM) explored using the service-oriented paradigm to develop distributed real-time applications and, together with Aveiro, devised an architecture to support dynamic service composition [46]. This kind of architecture can also be an alternative to provide some level of flexible resource management by supporting different profiles of each service, with different QoS and resource usage, and allowing a dynamic recomposition of the services/profiles involved in an application according to a predefined objective, e.g., maximize QoS of an application given the currently available resources, minimize the rsources needed by a set of applications, etc. The architecture is based on a global entity, called the composer, which, together with a QoS manager, decides when and how to recompose a given application. The real-time coordination of the composition changes in the distributed system is carried out with the FTT-SE protocol [18], [18b], [19]. A joint PhD thesis has been produced between UC3M and Aveiro. Also, UC3M is developing an improved version of HOLA-QoS architecture is being ported to real-time Java platforms to test its suitability for medium side embedded platforms.

Dual-rate switching control (UPC, Aveiro)
The group at UPC (affiliated to TUKL) and Aveiro developed an online adaptation strategy for feedback control loops that minimizes the respective resources used, both CPU and network. The strategy is based on using a pair of controller for each feedback loop, a nominal one and an adapted low bandwidth controller for periods of stationarity. The savings in resource usage are highly dependent on the dynamics of the application but can be explored by a flexible resource management framework (see joint publications below). Further joint work in a similar direction is being developed to carry out flexible resource management in the scope of feedback control applications themselves, with rate adaptation. to distribute bandwidth dynamically among several feedback control loops running on a single CPU (a joint publication has been submitted to a journal).

Rate adaptation in distributed control systems (Aveiro)
Rate adaptation allows varying the resources consumed by feedback control loops as required to carry out dynamic QoS management in such systems. The purpose is to maximize the number of entities, e.g., control loops, served by the system while keeping their QoS within acceptable ranges [13][14][15][16].

Feedback control of server-based real-time systems (UPC)
Feedback control of computing systems is an emerging area that aims at appliying control theory to computing systems in order to meet performance specifications regardless of vaying applications demands or unpredictable resource availability. This approach has been applied to server-based real-time systems, in order to automatically control the partial utilization of each server [20].

Overrun handling approaches for soft real-time periodic systems (U. of Catania)
In soft real-time environments featuring highly variable workload and resource requirements, tasks typically require probabilistic guarantees on meeting deadlines. When building the schedule for soft real-time applications it is therefore desirable avoiding worst-case assumptions on task execution times, in order to fully exploit the processor and increase system utilization as much as possible. To this end, “typical” task execution times (obtained either through previous measurements or from the knowledge of the execution time distributions and averages) can be used in the schedulability analysis instead of worst-case ones, which are very unlikely to occur. However, according to this approach, tasks can overrun, that is, the actual execution time required by some task instances can exceed the one assumed in the schedule. Due to overruns, it may happen that the system becomes overloaded, thus causing an unbounded number of deadline misses. As a result, a suitable overrun handling mechanism is needed in order not to jeopardize the system performance.

Traditional overrun handling approaches for real-time systems enforce some isolation property at the job or task level. This activity aimed at showing that by “relaxing” task isolation, it is possible to efficiently deal with overruns in soft real-time systems with highly variable task execution times, and proposed Randomized Dropping (RD), a novel overrun handling mechanism. RD is able to bound task overruns in a probabilistic manner, thus providing “soft” task isolation. It has been shown how to combine RD with priority-driven and rate-based scheduling algorithms, and how to analyze the resulting system [29]. Performance evaluation and comparison between simulation and analytical results were discussed.

Stochastic response time analysis of hybrid task sets in priority-driven soft real-time systems (U. of Catania)
Starting from the approach in [30], this activity led to the development of a more generic task model than the traditional one, which assumes each task being characterized by a fixed activation period and a Worst-Case Execution Time. According to the new model, a task is characterised by an Arrival Profile (AP) and an Execution Time Profile (ETP), both given by random variables with known distributions. To cope with the unbounded interference introduced by aperiodic tasks in the system, this activity targeted the encapsulation of sporadic and aperiodic tasks using server-based techniques. The calculus for obtaining the ETP of servers used to handle hybrid task sets was derived, and the results obtained are then applied to a previous stochastic analysis framework, making it applicable to a more general domain [31].Then, the calculation of stochastic Response Time Profiles (RTPs) of tasks hierarchically scheduled using server-based techniques in a stochastic analysis framework was addressed. Results are given in [32].

Server-based approaches for flexible scheduling in distributed embedded real-time systems (U. of Catania)
Previous work from the University of Catania proposing the adoption of the Total Bandwidth Server on the traditional Master/Slave (M/S) Bluetooth operating mode to support flexible scheduling of hybrid (periodic+aperiodic) messages on Bluetooth (BT) networks was extended towards a more flexible and efficient approach. Although successful in providing support to soft real-time traffic, the original approach based on TBS suffered from some limitations, in terms of scan cycle duration and waste of bandwidth, due to the M/S mode.This year research at the University of Catania therefore focused on a novel approach, called called M/S(EDF+CBS)+S/S(EDF), which combines deadline-aware scheduling with the Slave/Slave operating mode, thus overcoming the limitations introduced by the M/S mode. The approach handles both periodic and aperiodic traffic in a flexible and integrated way combining Earliest Deadline First (EDF) with the Constant Bandwidth Server (CBS) [34]. The new approach proved to be capable of increasing the amount of bandwidth available for both periodic and aperiodic transmissions, while keeping periodic traffic isolated from aperiodic traffic. The use of real-time scheduling techniques, i.e. EDF and CBS, allows periodic traffic deadlines to be met in the various scenarios examined, while in the same conditions the standard BT approach experiences significant periodic packet loss and deadline miss values [35].

The activity on the problem of integrating networked subsystems in resource constrained environments, carried out in collaboration with Malardalen University, Sweden, continued investigating in a joint publication the choice of the Server-CAN scheduler to use, examining the difference of the temporal guarantees provided by the usage of different Server-CAN schedulers [33]. Moreover, it was addressed how to configure the N-Server parameters for a subsystem and the integration process of subsystems using the Server-CAN concept was illustrated using a simple example of how to configure the Server-CAN framework for a “real” subsystem. The cost of using the Server-CAN framework, in terms of temporal degradation and overhead, was also assessed.

A middleware to combine energy-awareness and real-time support for distributed embedded systems used in monitoring applications (U. of Catania)
The main focus of this activity was a middleware, called an Aggregation layer, to be introduced between the MAC and Routing layers. Such a middleware, in combination with suitable routing techniques, is able to support soft real-time traffic while reducing the energy consumption of the wireless nodes, which have to work for long periods without the possibility of replacing their batteries. This layer mainly deals with reducing the amount of energy dissipated, while the Routing layer is entrusted with achieving the desired delivery speed, to support the transmission of soft real-time traffic. The proposed middleware has been implemented in the RTPAW framework [36].

Keynotes, Workshops, Tutorials

Keynote : Predictable response times in event-driven real-time systems.
Automotive 2006 - Security and Reliability in Automotive Systems
Stuttgart, October, 2006
M. González Harbour

Keynote: QoS-Based Resource Management
University of Thessaloniki
Thessaloniki, Greece– May 4th, 2007

The keynote talk was given in the context of a periodic departamental workshop open to researchers and university students.
The topic of the talk was resource management in embedded applications that are tolerant to some degree of flexibility and, therefore, admit QoS-based operation. The talk introduced the problems of these systems and the solutions ranging from the older contract-based approach, to centralised resource management, to distributed resource management, and higher level protocols for managing application execution.
Marisol Garcia Vals

Workshop NeRES: 2007 – Networks for Reconfigurable Embedded Systems
Aveiro, Portugal – April 2007

This workshop was organized in the scope of the activity Dynamic and Pervasive Networks (appears in more detail in that report) and it targeted discussing the network requirements to support reconfigurability in distributed embedded systems, as well as the adequacy of current protocols and middlewares for that purpose,. Several presentations were delivered focusing on middleware layers to support flexible resource management, which was one kind of reconfiguration that was addressed.

First Italian Workshop on Real-Time Embedded Systems
RETIS Lab, Scuola Superiore Sant’Anna, Pisa, July 2, 2007
Organizers: Giorgio Buttazzo, Giuseppe Lipari, Lucia Lo Bello
Objectives: Build an Italian community on real-time embedded systems and favor interactions between Italian industry and academic researchers.
Topics: real-time scheduling, operating systems, sensor networks, design methodologies
Results: The workshop attracted 16 universities and 12 industries working in the field, 28 short presentations were given and participants had time to meet, know each other and exchange information on their research interests. A second workshop is planned for next year.

WCET 2007: Worst Case Execution Time Analysis
RETIS Lab, Scuola Superiore Sant’Anna, Pisa, July 3, 2007
Organizers: Christine Rochange, TRACES group, IRIT, Toulouse, France
Objectives: Bring together people from academia, tool vendors and users in industry that are interested in all aspects of timing analysis for real-time systems.
Topics: Timing analysis, calculation methods for WCET, testing methods for WCET analysis, tools for timing analysis, compiler optimizations for worst-case paths.
Results: The workshop attracted 32 participants from different European countries and technical papers have been published in proceedings.

RTN 2007: Real-Time Networks
RETIS Lab, Scuola Superiore Sant’Anna, Pisa, July 3, 2007
Organizers: Ye-Qiong Song, LORIA, Nancy, France
Objectives: RTN focuses on the current technological challenges of developing communication infrastructures that are real-time, reliable, pervasive and interoperable.
Topics: Distributed systems, communication protocols, wireless sensor networks, mobile ad-hoc networks.
Results: The workshop attracted 30 participants from different European countries and technical papers have been published in proceedings.

OSPERT 2007: Operating Systems Platforms for Embedded Real-Time Applications
RETIS Lab, Scuola Superiore Sant’Anna, Pisa, July 3, 2007
Organizers: Scott A. Brandt, University of California, Santa Cruz, CA, USA and Kevin Elphinstone, University of New South Wales, Kensington, NSW, Australia.
Objectives: This workshop is intended as a forum for researchers and practitioners of RTOS to discuss the recent advances in RTOS technology and the challenges that lie ahead.
Topics: Support for component based development; Scalability, from very small scale embedded systems to full-fledged OSes; Real-Time on Linux; Interaction with reconfigurable hardware; Support for embedded multi-processor architectures; Security and fault tolerance for embedded real-time systems; Power-aware operating systems..
Results: The workshop attracted 18 participants from different European countries and technical papers have been published in proceedings.
URL: sbrandt/OSP...

Tutorial: “MARTE: A New Standard for Modeling and Analysis of Real-Time and Embedded Systems”
RETIS Lab, Scuola Superiore Sant’Anna, Pisa
July 3, 2007
The new standard provided by the Object Management Group, called the MARTE profile, addresses a broader scope than its predecessor (the SPT one). MARTE wants to tackle with all the activities of the two classical branches of the V cycle, i.e. modeling and validation& verification. Modeling capabilities have to ensure both hardware and software aspects of RTES in order to improve communication/exchange between developers. It has also to foster the construction of models that may be used to make quantitative analysis regarding hardware and software characteristics. The purpose of this tutorial is then to introduce the participants to the issues of model-driven development of RT/E applications and present how to use the new OMG standard for dealing with model-driven development of RT/E applications.

Publications Resulting from these Achievements

- [1] L. Risvanovic, D. Isovic, G. Fohler: “Integrated Global and Local Quality-of-Service Adaptation in Distributed, Heterogeneous Systems” 2007 IFIP International Conference on Embedded and Ubiquitous Computing (EUC-07)

- [2] Larisa Rizvanovic, Gerhard Fohler: “The matrix - a framework for real-time resource management for video streaming in networks of heterogenous devices.”The International Conference on Consumer Electronics 2007, Las Vegas, USA, January 2007

- [3] C. Otero-Perez, J v. EIndhoven, G. Fohler: “Execution-aware cache reservations for media applications”, submitted

- [4] Radu Dobrin, Gerhard Fohler: Handling Non-periodic Events Together with Complex Constrained Tasks in Distributed Real-Time Systems, 16th International Conference on Control Systems and Computer Science, Bucharest, Romania, May, 2007

- [5] Julio L. Medina, Patricia López, José M. Drake, François Terrier and Sèbastien Gerard. A Modeling Approach for the Timing Verification of COTS Components-based Distributed Hard Real-Time Systems. Workshop on Models and Analysis for Automotive Systems (RTSS2006-WMAAS), held in conjuntion with the RTSS-2006, Rio de Janeiro, Brasil, December 2006.

- [6] H. Espinoza, J. Medina, H. Dubois, F. Terrier, and S. Gerard. Towards a UML-Based Modeling Standard for Scedulability Analisys of Real-Time Systems. International Workshop on Modeling and Analysis of Real-Time Embedded Systems at MODELS’06, ISBN: 82-73-68-299-4, Genova (Italy), October 2006.

- [7] Nogueira, L., Pinho, L., "Iterative Refinement Approach for QoS-aware Service configuration", 5th IFIP Workshop on Distributed and Parallel Embedded Systems, Braga, Portugal. Oct. 2006, published in "From Model-Driven Design to Resource Management for Distributed Embedded Systems", B. Kleinjohann, L. Kleinjohann, R. Machado, C. Pereira, P. Thiagarajan (Eds.), IFIP Vol. 225, 2006, Springer, pp. 155-164, Oct. 2006.

- [8] Nogueira, L., Pinho, L., "Building Adaptable, QoS-aware Dependable Embedded Systems", in proceedings of the 3rd International Workshop on Dependable Embedded Systems (WDES 2006), Leeds, UK, pp. 72 - 78, Oct. 2006.

- [9] Nogueira, L., Pinho, L., "Capacity Sharing and Stealing in Dynamic Server-based Real-Time Systems", 21st IEEE International Parallel & Distributed Processing Symposium, Workshop on Parallel and Distributed Real-Time Systems, Long Beach, EUA, March 2007.

- [10] Nogueira, L., Pinho, L., "Handling Shared Resources and Precedence Constraints in Open Systems", WiP Session of the 19th Euromicro Conference on Real-Time Systems (ECRTS’07), Pisa, Italy, July 2007.
Universidade de Aveiro

- [11] L. Almeida, Sebastian Fischmeister, Madhukar Anand, Insup Lee. A Dynamic Scheduling Approach to Designing Flexible Safety-Critical Systems. EMSOFT 2007, 7th ACM Conference on Embedded Software. Salzburg, Austria. Sept 2007.

- [12] L. Almeida, P. Pedreiras, J. Ferreira, M. Calha, J. A. Fonseca, R. Marau, V. Silva, E. Martins. On-line QoS adaptation with the Flexible Time-Triggered (FTT) communication paradigm. in Handbook on Real-Time and Embedded Systems, I. Lee, J. Y. Leung, S. Son (ed). CRC Press, ISBN: 9781584886785, July 2007.

- [13] A. Antunes, P. Pedreiras, L. Almeida, A. Mota. Improving Operational Flexibility in Distributed Control Systems: The Dynamic Rate Adaptation Technique. Automazione e Strumentazione. Anno LV, Nº2, pp:90-97. VNU business publications, Italia, February, 2007.

- [14] A. Antunes, P. Pedreiras, L. Almeida, A. Mota. Dynamic Rate and Control Adaptation in Networked Control Systems. INDIN 2007, 5th IEEE Conference on Industrial Informatics, Vienna, Austria, July 2007.

- [15] A. Antunes, P. Pedreiras, L. Almeida, A. Mota. Dynamic Rate Adaptation: a method to improve operational flexibility in distributed control systems (poster). FeBID 2007, 2nd IEEE Workshop on Feedback Control Implementation and Design in Computing Systems and Networks. Munich, Germany, May 2007.

- [16] A. Antunes, P. Pedreiras, L. Almeida, A. Mota. Dynamic Rate Adaptation in Distributed Computer Control Systems. ANIPLA 2006 – International Congress of the Italian National Association for Automation, Rome, Italy, 13-15 November, 2006. (best paper award; selected for journal publication)

- [17] L. Almeida, P. Pedreiras, R. Marau. Traffic Scheduling Anomalies in Temporal Partitions. DIPES 2006, IFIP 5th Conference on Distributed and Parallel Embedded Systems, Braga, Portugal. October 2006.

- [18] I. Estévez Ayres, L. Almeida, M. García Valls, and P. Basanta Val. An Architecture to Support Dynamic Service Composition in Distributed Real-Time Systems. Accepted for publication in 10th IEEE International Symposium on Object/Component/Service-Oriented Real-Time Distributed Computing (ISORC 2007). Santorini Island, Greece. May 2007.

- [18b] I. Estévez Ayres. Techniques for flexible functional composition in distributed embedded systems. PhD Thesis. Universidad Carlos III de Madrid. Advisors: M. García-Valls and L. Almeida. February 2007.

- [19] P. Basanta Val, L. Almeida, M. García Valls, and I. Estévez Ayres. Towards a synchronous scheduling service on top of an unicast distributed real-time Java. Accepted for publication in 13th IEEE Real-time Systems Application Symposium (RTAS’07).

- [20] Manel Velasco and Pau Martí. A Preliminary Approach to Feedback Control of Server-based Real-Time Systems. In Poster Session, Second IEEE International Workshop on Feedback Control Implementation and Design in Computing Systems and Networks(FeBID07), Munich, Germany, May 2007.

- [21] Y. Chu and A. Burns, “Supporting Deliberative Real-Time AI Systems: A Fixed Priority Scheduling Approach”. In Proceedings of ECRTS07. pp. 259-268. July 2007.

- [22] G. M. A. Lima and A. Burns, “A Priority-based Consensus Protocol”. In Proceedings of the 25th Brazilian Symposium on Computer Networks and Distributed Systems (SBRC 2007). Volume 1, pp. 353-366, May 2007.

- [23] P. Emberson and I. Bate, “Minimising Task Migrations and Priority Changes In Mode Transitions”. In Proceedings of the 13th IEEE Real-Time And Embedded Technology And Applications Symposium (RTAS 07). pp. 158-167. April 2007.

- [24] A. Burns and T.-M. Lin, “An Engineering Process for the Verification of Real-Time Systems”. Formal Aspects of Computing. Volume 19, Number 1, pp. 111-136. March 2007.

- [25] R.I Davis and A. Burns, “Resource Sharing in Hierarchical Fixed Priority Pre-emptive Systems”, In Proceedings of the 27th IEEE Real-Time Systems Symposium, pp. 257-267, December 2006.

- [26] S. Baruah and A. Burns, “Sustainable Scheduling Analysis”, In Proceedings of the 27th IEEE Real-Time Systems Symposium, pp. 159-168, December 2006.

- [27] D. Kazakov and I. Bate, “Towards New Methods for Developing Real-Time Systems: Automatically Deriving Loop Bounds Using Machine Learning”. In Proceedings of the 11th IEEE International Conference on Emerging Technologies and Factory Automation, Sept. 2006.

- [28] N.C. Audsley, R. Gao and A. Patil, “Towards a File System Interface for Mobile Resources in Networked Embedded Systems”. In Proceedings of the 11th IEEE International Conference on Emerging Technologies and Factory Automation, Sept. 2006.

- [29] Lo Bello L., K.H. Kim. Overrun handling approaches for overload-prone soft real-time systems, accepted on Advances in Engineering Software, Elsevier.

- [30] J. Diaz, K. Kim, J. Lopez, L. Lo Bello, D. Garcia, C.Lee, S. Min, O. Mirabella, “An Exact Stochastic Analysis of Priority-Driven Periodic Real-Time Systems and Its Approximations”,The Handbook of Real-Time and Embedded Systems, Chapter 9, July 2007, Taylor and Francis Group, Boca Raton, FL, US.

- [31] G. A. Kaczynski, Lo Bello L., T. Nolte. Towards Stochastic Response-Time of Hierarchically Scheduled Real-Time Tasks. In: Proceedings of ETFA’06. The 11th IEEE International Conference on Emerging Technologies and Factory Automation. Sept. 20-22, 2006. (pp. 453-456). ISBN/ISSN: 1-4244-0681-1. PISCATAWAY, NJ: IEEE Industrial Electronics Society (US).

- [32] G.A. Kaczynski, L. Lo Bello, T. Nolte, “Deriving Exact Stochastic Response Times of Periodic Tasks in Hybrid Priority-driven Soft Real-time Systems”, ETFA 07, accepted.

- [33] T. Nolte, Lo Bello L., H. Hansson. Integration of networked subsystems in a resource constrained environment. In: Proceedings of ETFA06. The 11th IEEE International Conference on Emerging Technologies and Factory Automation. Sept. 20-22, 2006. (pp. 905-912). ISBN/ISSN: 1-4244-0681-1. PISCATAWAY, NJ: IEEE Industrial Electronics Society (US).

- [34] M. Collotta, L Lo Bello, O. Mirabella, “Deadline-Aware Scheduling Policies for Bluetooth Networks in Industrial Communications”.In Proceedings of the IEEE Second International Symposium on Industrial Embedded Systems - SIES’2007, Lisbon, Portugal, 4-6 July 2007.

- [35] M. Collotta, O. Mirabella, L. Lo Bello, “Comparison between RT scheduling techniques for Bluetooth Networks in DPCSs”. In Proceedings of the IEEE Second International Symposium on Industrial Embedded Systems - SIES’2007, Lisbon, Portugal, 4-6 July 2007.

- [36] E. Toscano, O. Mirabella, L. Lo Bello, “An Energy-Efficient Real-Time Communication Framework for Wireless Sensor Networks”, the 6th International Workshop on Real-Time Networks (RTN’07) in conjunction with the 19th Euromicro International Conference on Real-Time Systems (ECRTS’07), Pisa, June 2007.

- [37] Enrico Bini, Marco Di Natale, and Giorgio Buttazzo, "Sensitivity Analysis for Fixed-Priority Real-Time Systems", Real-Time Systems, to appear.

- [38] Giorgio Buttazzo and Enrico Bini, "Optimal Dimensioning of a Constant Bandwidth Server", IEEE Proc. of the 27th Real-Time Systems Symposium (RTSS 2006), Rio de Janeiro, Brasil, Dec. 2006.
- [39] Giorgio Buttazzo, "Why real-time computing?", 50th Int. Congress of ANIPLA on Methodologies for Emerging Technologies in Automation (ANIPLA 2006), Rome, Nov. 2006.
- [40] Giorgio Buttazzo, "Real-Time Scheduling and Resource Management", in The Handbook of Real-Time and Embedded Systems, Edited by Joseph Leung, Insup Lee, and Sang Son, CRC Press, 2006.



ARTIST2 Participants: Expertise and Roles

  • Activity Leader: Gerhard Fohler - Technische Universität Kaiserslautern
    resource management framework, video streaming, wireless networks
  • Team Leader: Giorgio Buttazzo - Scuola Superiore S. Anna
    scheduling mechanisms for temporal isolation
  • Team Leader: Michael Gonzalez Harbour - University of Cantabria
    end-to-end timing analysis of distributed systems
  • Team Leader: Luis Almeida - University of Aveiro
    bandwidth adaptation in networks
  • Team Leader: Eduardo Tovar - Polytechnic Institute of Porto
    communication protocols for wireless networks
  • Team Leader: Alejandro Alonso - UP Madrid
    adaptive techniques for QoS
  • Team Leader Alan Burns – University of York (UK)

    Role: fixed priority schemes and real-time languages

Affiliated Participants: Expertise and Roles

  • Team Leader: Marisol García-Valls - U. Carlos III, Madrid
    QoS management
  • Team Leader: Pau Martí - Universitat Politècnica de Catalunya
    real-time and control
  • Team Leader: Paolo Gai - Evidence, SME
    tools for resource management
  • Team Leader: Liesbeth Steffens - Philips Research
    consumer electronics, video streaming, multi resource management
  • Team Leader Ivo De Lotto – University of Pavia (Italy)
    Role: dynamic priority schemes
  • Team Leader Lucia Lo Bello – Univ. of Catania (Italy)
    Role: communication protocols and stochastic scheduling

(c) Artist Consortium, All Rights Reserved - 2006, 2007, 2008, 2009

Réalisation Axome - Création de sites Internet