Angelo Corsaro, PrismTech
Gerardo Pardo-Castellote, Real-Time Innovations

Products conforming to OMG’s Data Distribution Service (DDS) specification are used in a broad range of application domains. This short introductory overview briefly presents a few of the many areas where DDS is in use today.
 

Gerardo Pardo-Castellote, Real-Time Innovations

Today's system's requirements include distributed and net-centric concepts and capabilities. The challenge is to rapidly and accurately distribute large amounts of information to large numbers of nodes, over a variety of transports, account for application level Quality of Service (QoS) all while realizing a decoupling data-centric software architecture. Traditional point-to-point integration technologies and techniques simply don't work in these large distributed environments.

This tutorial introduces the OMG Data Distribution Service (DDS) and highlights the standard's unique capabilities that address and solve these real-time distributed system integration challenges. Specifically, the talk will include an introduction to publish/subscribe concepts, an overview of the specification, DDS data-modeling capabilities, and many of the application level QoS polices and their uses in real-world applications. The tutorial will conclude with a simple demonstration highlighting key concepts covered.
 

Angelo Corsaro, PrismTech

The DDS specification provides fine-grained control over the real-time behavior, dependability, and performance of DDS applications by means of a rich set of QoS Policies. The challenge for many DDS users is that the specifications explains very clearly how each QoS allows to control very specific aspects of data distribution yet it provides no hints on how different QoS should be composed to control complex properties such as the consistency model, or to impose end-to-end real-time scheduling decision. This half-day tutorial will fill this gap by providing attendees with (1) an explanation of how the various QoS compose, and (2) providing attendees with a series of QoS-composition Patters that can be used to control macro-properties of an application, such as the consistency model.
 

Chair: Julio Medina, Universidad de Cantabria

Toward a Safety Critical DDS Profile for Data Communication in Transportation Systems

Nawel Hamouche & Angelo Corsaro, PrismTech

A Communication-Based Train Control (CBTC) system is a continuous, automated control system for railways that ensures the safe operation of rail vehicles using data communication between various control entities that make up the system. CBTC systems present several challenges; CBTC systems have to comply with very strict certification requirements (up to CENELEC SIL4) that have hindered the adoption of proprietary as well as standard-based COTS, and different subsystems often use different data distribution technologies thus hindering seamless interoperability and data exchange. Among standard-based middleware technologies, the Data Distribution Service for Real-Time Systems (DDS) provides some very good abstractions for integrating CBTC subsystems as well as dealing with the communication among the most critical components. However, to make DDS a perfect fit there are many aspects that need to be addressed, including (1) DDS lacks a few features that are required to deal with communication between components running at a SIL4 level (the most critical level identified in the CENELEC standard). (2) DDS includes certain dynamic features, such as dynamic discovery and Request vs. Offered subscription matching that cannot be used in critical communication.

This presentation will summarize first the challenges of CBTC systems and will then introduce the minimal set of DDS features required by a CBTC system as well as the set of extensions required at a protocol level to better deal with high level of safety such as those required by the SIL4 components.
 

Gabriela Ciocarlie, Heidi Schubert & Rose Wahlin , Real-Time Innovations

A mixed-criticality system is one that must meet multiple assurance requirements, in particular safety and security requirements. Ideally, such a system could be constructed with components that have been individually certified to meet safety or security requirements relevant to their function. The challenge is that to be effective components need to communicate with each other. In this case, the safety and security properties of a certified component can be affected by changes in other components that it is linked to. In this paper we propose a method to alleviate this problem, by moving from a component-interaction model to a data-centric model, and a specific instantiation of it based on the Object Management Group (OMG) Data Distribution Services (DDS) standard middleware (which defines and disseminates data) and the Wind River VxWorks MILS platform (which ensures component separation). The challenges encountered and the lessons learned are also presented, along with recommendations that could be used to guide a future Safety and Security profile for DDS specification.
 

Real time Data Distribution for Airborne Systems

Ramzi Karoui & Angelo Corsaro, PrismTech

Historically avionics application used to rely on physical partitioning in order to address the safety, availability and reconfigurability requirements. This approach, which is still used, leads to under-utilization of computational resources (especially in a world dominated by multi-core processors) and more importantly adversely impacts the space and weight constraints of an avionic system. To address and overcome these limitations, avionics manufacturers have introduced a new architectural approach to building avionics systems - the Integrated Modular Avionics (IMA). This promises better resource usage, higher reconfiguration capability and better extensibility. Furthermore, IMA fosters platform independence, time- and space- partitioning, and incremental software updates and certification.

Parts of the IMA requirements are addressed by middleware technologies that promote and support real-time decoupling architectures, such as the OMG Data Distribution Service for Real-Time Systems (DDS). In this presentation we will (1) demonstrate the close architectural fit existing between DDS and IMA, (2) enumerate the advantages that the avionic industry could take out of DDS and (3) detail an integration use case between OpenspliceDDS and an ARINC-653 compliant Real-time OS. Finally, we will provide some preliminary performance evaluations of OpenSplice DDS running on an ARINC-653 platform.
 

Angelo Corsaro, PrismTech
Gerardo Pardo-Castellote, Real-Time Innovations
Paul Giangarra, IBM
Clark Tucker, Twin Oaks
Ken Rode, Gallium Visual Systems

Five vendors of DDS-compliant middleware give a live interoperability demonstration.
 

Chair: Gerardo Pardo-Castellote, Real-Time Innovations

An Extensible DDS-based Monitoring and Intrusion Detection System

Fernando Garcia-Aranda & Juan M. Lopez-Soler
Dept of Signals, Telematics & Communications, University of Granada
Javier Sanchez-Monedero, Department of Computer Science & Numerical Analysis, University of Cordoba

Department of Computer Science & Numerical Analysis, University of Cordoba
As centralized computing models are being superseded by more distributed computer environments, supervising application, system and network health has become a crucial task in network management. Currently, there are plenty of systems providing a partial solution to this problem (e.g. intrusion detection systems (IDS) and system monitoring tools), but each separately cannot provide the complete view required to understand and react quickly.

In this context we propose Cave Canem, a framework for the development of a "complete" operational picture for the combination of multiple kinds of information (e.g. performance, system alerts, intrusion detection), remaining open so that new sensors and applications can be mixed in. This open approach, separates the Information Model from the Information Distribution mechanisms, using the OMG Data Distribution Service (DDS) standard as the fundamental information-sharing model.

Our framework exploits the benefits of the data-centric publish-subscribe (DCPS) paradigm provided by the DDS middleware, which eases the deployment and maintainability of large monitoring scenarios. It also takes advantage of some of the DDS features, such as content filtering and QoS settings (reliability, transport priorities, etc.) which ensure that the entire system status can be efficiently distributed according to the subscriber interests.
In the presentation we will purpose a motivating example for highlighting our design goals, describe the architecture and the implementation details of our solution, and demonstrate a validated prototype.
 

Use of OMG DDS in Simulation; A new Way of developing Real Time Distributed Simulation

Jose M Lopez-Rodriguez, NEXTEL Engineering Systems

HLA is the most common standard applied to distributed simulation. HLA is defined under IEEE Standard 1516 and it is also a NATO standard (STANAG 4603). HLA successfully enabled interoperability between simulation systems, mainly in military projects, because it has provided a way to do distributed simulation and execute virtual exercises between simulation systems of different NATO countries and providers. Indeed HLA is the mandated standard for the principal distributed simulation projects in NATO, like Allied Command Transformation (ACT) project Snow Leopard. But HLA has not achieved a greater challenge: open the internal software architecture of the simulation systems, in order to achieve reusability of simulations models and simulator components. This paper describes the use in Simulation of another international real-time publish-subscribe standard: OMG DDS (Data Distribution Service). DDS-specific features for management and data distribution in large critical distributed systems and federation of systems are very suitable for real time distributed simulations. The paper explains how these features create an excellent foundation for an open real time simulation architecture.

The advantages of this new simulation architecture will be exposed and it will show that DDS is the best companion for HLA in Simulation and its use enables great savings in the costs and delivery times of the simulation projects. We include case studies showing how this new DDS-based architecture has been successfully applied to real time virtual and engineering simulators.
 

Scientific Applications of Data Distribution Service

Svetlana Shasharina, Nanbor Wang, Rooparani Pundaleeka, James Matykiewicz and Stephen Goldhaber
Tech-X Corporation

This talk describes our efforts in bringing DDS to some scientific applications. In particular, we will review astronomical data processing applications and present our approach to accommodate their workflows using DDS. Next, we will discuss utilizing DDS to provide the messaging bus for a unified high-level application environment for particle accelerator control systems. We will also review the sub-projects we are working on in support of these applications, including the application of the new C++ mapping for DDS, the secure communication channel, and Python DDS API support.
 

MARTE to MAST: Schedulability Analysis of Real-time Distributed Systems with UML & MARTE

Julio Medina, Universidad de Cantabria

This presentation describes the design of, implementation of , and some initial experiences with a tool that helps to construct the schedulability analysis models used as input to the MAST set of tools extracted from UML models handled by the eclipse UML2 plug-in. The analysis models correspond to concurrent, distributed real-time systems that are formalized in UML and whose non-functional properties are annotated with the extensions proposed by the UML Profile for MARTE; the OMG standard profile for the modelling and analysis of real-time and embedded systems. The initial experiences performed with the modeling elements that have been selected to support the underlying analysis modeling methodology have led to the rising of concrete issues, which are to be sent to the OMG for the improvement of the standard. The issues found, as well as the prospective solutions, which will be discussed in the current MARTE 1.3 Revision Task Force will be also briefly described in the presentation.
 

Antonio Bovenzi, Sergio Celentano & Christian Esposito, Dipartimento di Informatica e Sistemistica, Universita degli Studi di Napoli "Federico II"
Gabriella Carrozza, Aniello Napolitano & Antonio Strano, SESM S.c.a.r.l.

Several industrial projects aiming at realizing the future generation of critical systems called Large-scale Complex Critical Infrastructures (LCCIs), which require a scalable and robust data dissemination infrastructure. Due to the well known decoupling properties exhibited by DDS-complaint products, which enforce the scalability achievable by the middleware, these have been increasingly using for architecting LCCIs. This is witnessed, for example, by the fact that they are the highly rated candidates to realize the data dissemination infrastructure within the SESAR (Single European Sky ATM Research) research and development program. This aims to develop a European wide air traffic management system to face the growing demand envisioned for the next 20 years. Such a scenario poses serious dependability requirements, which in turn require DDS to realize reliable data dissemination. Hence, assessing the supported reliability degree, and proposing reliability improvement strategies, becomes crucial and requires a clear understanding of how DDS-complaint products behave when failures occur.

This presentation illustrates an approach for the evaluation of the robustness of DDS-compliant products, based on a fault injection tool that automatically submits faulty values to API functions. The approach has been applied to a real industrial prototype used as SESAR pilot system, named SWIM-BOX, developed by SESM in the framework of SESAR research activities. Experiments have been conducting by injecting a set of invalid inputs into a DDS compliant implementation and by monitoring middleware reactions. Results have been collected and evaluated against well-known robustness metrics in order to disclose some important hazards in the middleware usage.
 

CCM and its New Extensions

Johnny Willemsen & Marcel Smit, Remedy IT

CCM is available as a component model for over a decade now. Recently there has been renewed interest in CCM, which has lead to the standardization of DDS4CCM and a new OMG RFP for AMI4CCM. First we will give a brief introduction to CCM, focusing on the development of CCM components. We than describe the new DDS4CCM specification, which provides a way to integrate DDS smoothly into CCM. To integrate asynchronous two-way communication into CCM, we proposed an AMI4CCM extension to the OMG. We will give an insight into our ideas for this extension. We will finish with some ideas on how a revised CCM specification would make CCM more extensible and usable in non-CORBA environments.
 

William R. Otte, Vanderbilt University

Large-scale distributed real-time and embedded (DRE) systems often have stringent Quality of Service (QoS) requirements, such as their end-to-end deadline, throughput, and jitter. These requirements are nominally an inherent complexity that substantially complicates the development of these applications. Solutions such as Real-Time CORBA (RTCORBA) have simplified the process of portably providing resolution to these requirements, but their use is hobbled by the tight binding that they must have to the application business logic making it difficult to adapt to changing requirements. The use of a container model, such as the CORBA Component Model (CCM), which provides a clear separation of concerns between the application business logic and client/server configuration, along with a well-defined deployment and configuration system, such as the Deployment and Configuration Specification (D&C), provide a vehicle to alleviate this accidental complexity of RTCORBA. This presentation will provide an overview of a new RT-CCM mapping that takes advantage of the recent developments in the Deployment and Configuration Engine (DAnCE), an implementation of the D&C standard, which allows for the explicit installation, instantiation, and configuration of component server, containers in addition to the previously present functionality that allowed it to configure only components and homes.
 

Akshay Dabholkar, Vanderbilt University

Component-based distributed real-time and embedded systems are often deployed in environments that simultaneously require non-functional properties; timeliness, fault-tolerance, and load balancing capabilities. In order to safely support these quality of service (QoS) properties simultaneously, component middleware developers have to constantly face obstacles in the design and implementation of the component containers due to the general-purpose nature of the middleware implementations. As a result, the developers have to make compromises, such as creating custom containers to support each individual QoS. Such approaches not only force unwanted rewriting of application code, but are also cumbersome to maintain. A possible solution is a generic container implementation that has safe, specialized support for multiple QoS properties, so that no two QoS interfere with one another at design-time or at run-time. We propose a dynamic container design that can be easily specialized to enable applications to seamlessly and safely adapt to particular execution environments by changing their QoS without requiring rewrite. We demonstrate how a generic container can be realized in LwCCM.
 

High-level Programming of DDS Systems

Alejandro de Campos Ruiz, Gerardo Pardo-Castellote, GianPiero Napoli, Fernando Crespo-Sanchez & Javier Sanchez Monedero, Real-Time Innovations

Analysis of the DDS API and the source code of typical DDS applications reveals three types of application code: (a) DDS Configuration and Setup code, (b) DDS Action code, and (c) Domain-Specific code that does not involve the use of the DDS API. If we discount the domain-specific code (c), which for the most part is independent of the middleware, we can see that in typical applications 80% of the DDS code is setup code. Moreover looking at the DDS API itself, we also see that about 50% of the API itself supports configuration and setup (e.g. QoS policy definitions, factory operations to construct entities) and only the remaining 50% supports action code.

These observations suggest there can be tremendous benefits to providing a mechanism that allows application developers to configure the system declaratively and, in doing so, avoid the need to explicitly learn and use the Configuration and Setup part of the DDS API.
We will present recent research on a new approach to developing DDS applications where the configuration & setup code is replaced with declarative statements written in XML and interpreted at run-time. We suggest extensions to the DDS standard API to allow the high-level programming approach to work on top of any DDS-compliant implementation.
 

Jaime Martin-Losa, eProsima
Gerardo Pardo-Castellote, Real-Time Innovations

The OMG Data-Distribution Service (DDS) is quickly becoming the technology of choice for communication and integration of field-deployed systems, such us, battlefield vehicles, UAS and UGS, ship-to-shore communications, etc. A common aspect of all these systems is their dependency on a "disadvantaged" communications infrastructure (e.g. field radios and acoustic modems) with limited bandwidth, higher loss rates, and intermittent connectivity.
DDS is attractive in this environment because of its efficient real-time publish-subscribe (DDS-RTPS) protocol, compact binary data representation, built-in services such as data-caching, content-filtering, persistence that allow applications to gracefully survive the loss of connectivity, and support for QoS, such a LatencyBudget, TimeBaseFiltering, Lifespan, Liveliness that allow applications to effectively adapt and use the available bandwidth.
However there are challenges, especially in situations with very limited bandwidth, which could be lower than 4Kbps in ground radios, and less than 200bps on underwater acoustic modems. In this talk we will present recent research on the use of DDS on field radios, describing the lessons learned, experimental data, and performance results obtained before and after suggested improvements to the DDS discovery and DDS-RTPS interoperability wire protocol.
 

Stephen Rhee, Richard Warren & Gerardo Pardo-Castellote, Real-Time Innovations

The increase of CPU power, drop in price, and widespread availability of IP-based networks is making it possible for millions of devices to be constantly connected, allowing large networks of sensors and devices to be monitored and processed in real-time. Systems such as power meters, home telecom appliances, and industrial control equipment are being integrated into vast networks of tens or hundreds of millions of devices that span the word. Because there is no direct human involvement in the operation of these networks, the amount and frequency of data they handle is limited only by the hardware. Deploying an efficient architecture is therefore critical to realizing the potential of the infrastructure.

The OMG DDS is emerging as a potential candidate infrastructure for such networks. However despite its scalability, ease of deployment and efficient data-centric publish-subscribe model, there are significant challenges to using DDS in this way, such as the practicalities of routing information between specific devices in a network of several million, and the scaling problems of hub-and-spoke configurations where several million clients feed data to a small number of supervisory nodes.

This presentation illustrates a proposed solution that allows DDS to scale to networks which over 1 Million nodes while addressing these problems. Simulation results are presented to illustrate the effectiveness of our approach.
 

Chair: Andrew Watson, OMG