Angelo Corsaro, Product Strategy & Marketing Manager, PrismTech
Gerardo Pardo-Castellote, CTO, 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.

UML Profile for Modeling and Analysis of Real-time and Embedded Systems (MARTE) Tutorial

Chokri Mraidha, Commissariat a l Energie Atomique-CEA/LIST
Bran Selic, President, Malina Software Corporation

Experience using "vanilla" UML in the development of time-critical and resource-critical systems shows that, while useful, it nevertheless lacks some key modelling elements needed in this domain, including quantifiable notions of time and resources, and specific real-time constructs such as task and semaphore. In addition, some aspects of the language's semantics are also not rigorously-enough defined for RT/E systems development. Fortunately (and contrary to an often-expressed opinion) UML's extensibility features can be used to address these issues, allowing the definition of a language profile that standardises the UML representation of RT/E concepts. The first such standard was the UML Profile for Schedulability, Performance and Time (SPT), adopted by OMG in 2002, which mainly focused on model-based RT analysis (and especially rate-monotonic analysis for schedulability analysis and layered queuing analysis for performance analysis). OMG's has now introduced a new second-generation UML RT/E profile to supercede SPT – the UML profile for Modelling and Analysis of Real-time and Embedded Systems, or MARTE for short. MARTE has a broader scope than its predecessor, and aims to tackle all the activities of both branches of the classical V cycle – modeling, and validation& verification. Its modelling capabilities support both hardware and software aspects of RT/E systems, in order to improve communication between developers, and foster the construction of models that can be used in quantitative analysis of a design's hardware and software characteristics. This tutorial covers both the foundations of model-driven development of RT/E systems, and the use of MARTE for model-driven development.

Angelo Corsaro, Product Strategy & Marketing Manager, PrismTech
Gerardo Pardo-Castellote, CTO, Real-Time Innovations

The continuation of this all-day tutorial.
 

Salvatore (Sam) Mancarella, Chief Technology Officer, Sparx Systems Pty Ltd. Angelo Corsaro, Product Strategy & Marketing Manager, PrismTech

The OMG Data Distribution Service is a widely adopted standard for real-time data-centric publish/subscribe services. It defines two levels of technology. The first level - the Data-Centric Publish Subscribe (DCPS) - contains the entities to provide data distribution capabilities with Quality of Service policies to govern its dissemination. The second level - the Data Local Reconstruction Layer (DLRL) - enables DDS to make this data available to the user as a cache of aggregate, object-oriented classes. The challenge for DDS application developers lies in applying well-known architecture and development practices to manage design complexity, verify design integration, and facilitate design reuse. The UML Profile for DDS provides modeling constructs, for both DCPS and DLRL that enable the use of Model-Driven Development and Model-Driven Architecture practices to meet these challenges for large-scale projects.
 

Program Chair: Andrew Watson, Object Management Group
 

Chair: Jaiganesh Balasubramanian, Vanderbilt University

The OMG Data Distribution Service is a standard for real-time data-centric publish/subscribe that has been widely adopted in deployments characterized by relatively homogeneous and controlled network environment. For instance, there are many mission- and business-critical deployments that span several LANs containing hundreds of nodes, and others within single LANs interconnected over a metropolitan area. Papers in this session examine the challenges of using DDS in Wide-Area Network (WAN) environments, including difficulties dealing with Firewalls and NATs traversal, limitations imposed by the current standard discovery protocol (both in terms of scalability to tens of thousands of nodes and its dependency on IP multicast), the suboptimal mechanism currently used for data distribution over unicast, and the lack of data compression.

Extending the Data Distribution Service for Internet-Scale Applications

Angelo Corsaro, Product Strategy & Marketing Manager, PrismTech
Sara Tucci-Piergiovanni, Researcher, University of Rome "Sapienza"

This presentation will provide a survey and comparative analysis of the techniques used by successful P2P technologies such as Skype, KaZaA, Gnutella, etc. Then, capitalizing from the experience of successful and scalable services such as KaZaA, it will provide an description of how the current OMG DDS standard can be extended to properly scale to the Internet scale while addressing the current issues associated with discovery, communication for nodes behind NAT/Firewalls, and data distribution efficiency in absence of multicast.

Deploying DDS on a WAN and the GIG: The DDS Router

Gerardo Pardo-Castellote, CTO, Real-Time Innovations
Fernando Sanchez, Principal Engineer, Real-Time Innovations
Jose M. Lopez-Vega, Ph.D student, University of Granada, Spain

This presentation introduces new research perform by RTI in collaboration with the University of Granada, Spain towards the development of a DDS-Router service that could be used to solve DDS's WAN deployment issues. Specifically we will discuss approaches to scaling to over 10000 nodes in a transparent manner, techniques to bridge data between different DDS Global Data Spaces, mechanisms to allow clients behind a Firewall or NAT to join a DDS Global Data Space, and required extensions to the DDS specification and the DDS-RTPS Interoperability protocol to enable these solutions to work across multiple DDS implementations. The presentation will conclude with a set of recommendations that could be used as guidance for the evolution and extension of the DDS standards family.

Chair: Andrew Watson, Object Management Group

Papers in this session cover various aspects of integrating DDS with other technologies.

From the Tactical Edge to the Enterprise: Integrating DDS and JMS

Rick Warren, Principal Software Engineer, RTI

The DDS and JMS specifications offer alternative visions of publish-subscribe communication and have traditionally been deployed in different capacities; DDS provides a data distribution model based on strong typing and flexible QoS-based control over data delivery, while JMS provides a traditional messaging model based on loosely structured data and emphasizes a simple and portable programming model. Next-generation service-oriented architectures will incorporate both of these domains and DDS and JMS each play their own role within larger systems of systems. In this talk, RTI will present the architecture of a DDS-interoperable, JMS-compliant messaging system based on conclusions from its research in the above areas and discuss developing – and potential – standards at the OMG and elsewhere and their relevance for edge-to-enterprise integration.

Accessing real-time DDS data from web-based clients redux: The WS-DDS Service

Gerardo Pardo-Castellote, Ph.D., Chief Technology Officer, Real-Time Innovations
Andrea Iannetti, Software Engineer, Real-Time Innovations
Fabrizio Bertocci, Principal Software Engineer, Real-Time Innovations

The Data-Distribution Service (DDS) is a widely used standard for distributing real-time publish-subscribe in performance and QoS-sensitive systems such as Combat Management Systems, Air Traffic Management and SCADA systems. As a consequence a lot of the “tactical” and “real-time” data in these systems “lives” in the DDS Global Data Space. A Web-Services API to the DDS Global Data-Space has been recognized as a key enabling technology that would facilitate access to this real-time data from non-realtime or enterprise applications, and enable “pure” web-applications whether server-based using scripting and CGI technologies, or client-bases using JavaScript or AJAX technologies to access the Global Data Space. This presentation builds upon the ideas presented during last year's workshop and introduces recent research towards developing such Web-Services API to DDS, as well as the lessons learned from implementing a prototype WS-DDS Service component.

Chair: Julio Medina, Universidad de Cantabria

This session offers presentations of experience reports from implementers and users of a number of real-time specifications.

Meeting End-to-End QoS Requirements

Gair D. Brown, Naval Surface Warfare Center
Richard Lescroart, Senior Principle Engineer, Raytheon IDS Seapower Capability Center
John Robert, Senior Member of Technical Staff, Software Engineering Institute - CMU
Antonio Samuel, Naval Surface Warfare Center

System development teams that use standardized middleware products to achieve end-to-end Quality of Service (QoS) requirements must overcome several challenges to successfully integrate these products and meet the overall system QoS Requirements. Common challenges include identifying the critical QoS requirements, selecting middleware products that support these QoS requirements and integrating the products into the system architecture. Some middleware products have aggressively pursued capabilities to utilize real-time specifications and underlying OS scheduling mechanisms to target low latency and high determinism needs. However, additional challenges include identifying the corrective actions to be taken when QoS requirements cannot be met and working with vendors to align the needed product improvements with marketplace needs. This presentation shares the experience from a system development team to meet end-to-end QoS requirements including discussion of the challenges and lessons learned. The presentation includes multiple perspectives from the team including authors from the government, industry and the Software Engineering Institute.

Achieving DDS Interoperability

Clark Tucker, President, Twin Oaks Computing
 

Twin Oaks Computing participated in a multi-vendor DDS interoperability demonstration during the March 2009 OMG Technical Meeting in Arlington, VA, USA. In this presentation, we will describe the events leading up to the demonstration and discuss how we were able to achieve success. We will cover challenges faced, lessons learned, and touch on related future work. Finally, we will provide a few examples of items that we believe could be improved or added in future versions of the standards. Twin Oaks' position, as a vendor not involved with the development of these standards, provides a unique perspective on the DDS API and RTPS wire-protocol.

Chair: Andrew Watson, Object Management Group

Papers in this session examine new developments in real-time communication protocols.

Leveraging DDS-RTPS Wire Interoperability Protocol: Towards a canonical tool-set for run-time interaction with the Global Data Space

Fabrizio Bertocci, Principal Software Engineer, Real-Time Innovations
Alejandro Campos, Software Engineer, Real-Time Innovations
Gerardo Pardo-Castellote, Chief Technology Officer, Real-Time Innovations

This presentation exposes recent research in DDS Global-Data Space tools that leverage standard display models such as Microsoft Excel’s; it also explores the different categories of DDS run-time tools drawing requirements and specific examples from COTS tools commonly used to interact with data. We will analyse requirements, capabilities and use-cases for DDS run-time tools; how meta-data should be captured and displayed; dependencies on specific features of DDS and the DDS-RTPS interoperability protocol and gaps in the standards that might prevent some requirements from being met.

Adaptive Middleware And Network Transports for Real-time Event Stream Processing Applications

Joe Hoffert, Graduate Research Assistant, Vanderbilt University
Dr. Douglas C. Schmidt, Professor of Computer Science, Vanderbilt University

Various transport protocols can affect the performance of real-time event notification systems particularly in dynamic environments. Systematic evaluations are generally lacking, however, of transport protocols integrated with real-time event notification middleware. We have integrated the Adaptive Network Transport (ANT) framework with two implementations of the Data Distribution Service (DDS). The integration with the DDS implementations is not meant for performance comparisons between DDS products, but to showcase different integration approaches and also to generalize experimental results of the impact of transport protocols across DDS implementations. ANT includes support for the Ricochet transport protocol, a scalable reliable multicast protocol that combines high data rates with strong probabilistic delivery guarantees. We will characterize the performance of various transport protocols, including Ricochet, the standard DDS Interoperability (DDSI) protocol, TCP, UDP, and IP multicast. We also will present experimental results that quantify the trade-offs of the various transport protocols in a range of representative environments using DDS.

Comparing and Contrasting XMPP and DDS

Angelo Corsaro, Product Strategy & Marketing Manager, PrismTech
Gerardo Pardo-Castellote, Chief Technology Officer, Real-Time Innovations

The Extensible Messaging and Presence Protocol (XMPP) is an open, XML-based, protocol aimed at near-real-time instant messaging (IM) and presence information. This protocol, which has recently been extended to support some generic pub/sub functionalities has been in some instances compared with DDS, however to date a complete and systematic evaluation of the key similarities and differences between the two technologies is still missing. This presentation will provide a crisp side-by-side comparison of the XMPP and DDS technology explaining how their differences make them suitable for different use cases, and includes some comparative performance evaluation contrasting the efficiency and scalability of leading XMPP implementations with that of leading DDS implementations.

SESSION 5

Chair: Victor Giddings, Objective Interface Systems

Dependability and assurance continue to be issues in real-time and embedded distributed systems middleware. It is difficult to accommodate the wide range of desired application-specific policies and mechanisms in a standard set of features of products that reside “between” the application and the underlying Operating System and communications infrastructure. The necessary performance and complexity costs of these are also of greater concerns in often resource-constrained embedded systems or in systems with real-time responsive constraints. The presentations in this session will explore the frontiers in a number of directions.

MyCCM-HI, an Experimental Framework Dedicated to Critical and Reconfigurable Real-Time Systems

Etienne Borde, Thales Land & Joint Systems
Olivier Hachet, Thales Land & Joint Systems
Gregory Haïk, Thales Land & Joint Systems

We present our experience of using OMG's Lightweight CCM and D&C standards in the design of mission-critical and reconfigurable real-time embedded systems. Our open-source framework, MyCCM High Integrity, is a custom C-language implementation of the OMG Lightweight CCM specification, with add-ons and limitations dedicated to critical and reconfigurable systems. This framework also incorporates a meta-model that extends and limits D&C concepts to address our specific objectives. We will discuss these objectives and corresponding modifications in the D&C concepts. The implementation of our framework runtime also uses AADL (Architecture Analysis and Design Language) execution semantics. We will also present our views on opportunities for reusing this work in the definition of a component-based standard dedicated to critical and reconfigurable systems.

Resource-Aware Deployment and Configuration of Real-time Fault-Tolerant Middleware

Jaiganesh Balasubramanian, Graduate Student, Vanderbilt University
Dr. Douglas C. Schmidt, Professor of Computer Science, Vanderbilt University

This presentation describes how our quality-of-service (QoS)-enabled middleware (which is built upon RT-CCM and FT-CORBA) provides a holistic and automated solution to fault-tolerance and real-time deployment & configuration (D&C) through two novel capabilities. First, our QoS-enabled middleware provides a deployment-time allocation and scheduling algorithm that maps passively replicated application components to appropriate hosts subject to their soft real-time requirements and determines the failover order of application replicas based on their worst-case state synchronization delays. Second, our QoS-enabled middleware's model-driven D&C engine deploys and configures replicas for each application and provides resource-aware failover and delay-bounded state synchronization between backup and primary replicas.

Ariel Salomon, Senior Software Engineer, Real-Time Innovations
Gerardo Pardo-Castellote, Chief Technology Officer, Real-Time Innovations

When DDS is used in complex distributed systems-of-systems, numerous security concerns may arise. These include ensuring availability of the system, resistance of the system to un-trusted network traffic, and confidentiality of sensitive data. As systems are increasingly interconnected, it is crucial that security concerns be considered early in the design process. A solution that can address some of these problems is the use of an operating system with support for security policy control. One of the most versatile OS approaches to security is that of type enforcement, the model used by Security-Enhanced Linux (SELinux). We present an example system showing the use of SELinux to enforce security constraints. The tradeoffs involved in addressing the identified security risks will be discussed, which include issues of performance and system complexity. Testing confirms that the SELinux implementation imposes a minimal performance penalty. Encrypted data traffic, on the other hand, can result in significant overhead. It is also important to ensure that security policies are maintainable by operations staff once the system is deployed. A set of recommended uses of SELinux in DDS systems will be discussed.

Towards a Safety Critical profile for DDS

Rose Wahlin, Senior Applications Engineer, Real-Time Innovations
Gordon Hunt, Chief Applications Engineer, Real-Time Innovations

The use of DDS has gained widespread adoption in command-and-control applications due to its support of the publish-subscribe communications model typically found in such systems. DDS has been field-proven in extremely complex systems consisting of thousands of endpoints with high data rates and stringent latency budgets. Nevertheless, DDS has yet to gain traction in avionics applications due primarily to their certification requirements. However, the same data-centric publish-subscribe communication and performance drivers that make DDS so attractive in the bit tactical systems also exist in avionics systems. This presentation will discuss research and prototype work towards the definition of a "safety critical" profile of DDS. We will describe an analysis of the DDS minimum profile features and relative code complexity oriented towards identifying which features could be implementable within the size constraints imposed by a "safety-certifiable" version of DDS. We identify a candidate subset of the minimum profile that has been found to meet the requirements of many avionics applications, and the lessons learned from implementing a "small footprint" version of DDS. Issues regarding the interoperability between a safety-critical implementation of DDS and a full-featured DDS implementation will be presented, along with the Quality of Service necessary and proposed for a small-footprint DDS for safety critical applications.

Bruce Powel Douglass, Chief Evangelist, IBM Rational Software

The UML is a powerful modeling language but has not been widely used to represent or perform safety analysis. The most common safety analysis technique is known as Fault Tree Analysis (FTA). An FTA represents how normal conditions and events combine with faults to result in hazards during the deployment or usage of a system. This talk presents a profile for the UML that enables architects, systems engineers, to directly draw and analyze FTA diagrams.

New standards directions

Chair: Andrew Watson, Object Management Group

Papers in this session describe new work aimed at evolving OMG standards to keep pace with emerging technology.

MOSIC - A project to design and implement an open-source version of the OMG's new DDS for Lightweight CCM Specification

Andrew Foster, Product Manager, PrismTech

DDS has experienced rapid take-up since its adoption, and is proving to be excellent technology for building Distributed Real-time Systems (DRE) due to its high performance, flexibility and the rich set QoS polices it provides the application developer. Increasingly however, users are looking to leverage the benefits provided by service centric technologies such as the CORBA Component Model (CCM) in combination with the complementary data centric approach of DDS to develop the next generation of mission-critical information systems. This presentation will discuss the work carried out as part of MOSIC, a French regional government funded research and development project whose purpose is to develop an open-source reference implementation of an integrated CCM and DDS solution. Partners in the project include Thales, PrismTech, OBEO, and ENST. The objectives of the MOSIC project include proposing a coherent architecture supporting the integration of CCM components and DDS, and standardize the results of this work at the OMG. The presentation will discuss the current status of the project and review its roadmap and timescales.

The Design and Implementation of a Data Caching Extension for the OMG DDS

Angelo Corsaro, Product Strategy & Marketing Manager, PrismTech

Data Caching Technologies are becoming a key element for several different classes of applications, allowing frequently-used data to be stored closer to the application, possibly to avoid expensive access to a DBMS, complex re-computation, or simply making it possible to capture a working set out of a far bigger data set that might even exceed the physical memory available on any single machine of the given distributed system. However, many caching designs are based on centralized or federated architectures that pose serious scalability challenges or introduce single points of failure, and thus are not suitable for real-time or high-performance systems. We’ve designed and implemented a caching extension for the OMG DDS that addresses this shortcoming. As well as leveraging DDS's strengths to provide scalability and performance, our design minimizes cache access time, and provides a flexible pre-fetching mechanism to limit cold start misses. This presentation will provide a detailed explanation of our architecture, which we hope will motivate future extensions of the DDS standard.

Towards a New IDL to Ada2005 Mapping

Julio Medina, Universidad de Cantabria
Patricia López, Universidad de Cantabria
José Mª Drake, Universidad de Cantabria
Pablo Pacheco, Universidad de Cantabria

New capabilities introduced in the Ada 2005 language standard provide opportunities for a simpler but powerful CORBA IDL mapping. In particular, the introduction of interfaces as a native concept in Ada and provision of single and multiple inheritance could simplify both code generation from IDL and reverse engineering processes, while the new Ada containers library could help map sequences in CORBA. This presentation starts by briefly highlighting Ada 2005's new features, particularly those that may lead to a new and simpler mapping between Ada and IDL. We will then propose concrete technical directions to get an efficient mapping, following sound object oriented practices. We have tested some of these proposals in a prototype tool built using configurable templates and XML technologies. Finally, we will outline an action list for the preparation of a full new IDL to Ada mapping, with the objective of moving towards an OMG standard.

Chair: Andrew Watson, OMG