Article by: Jeffrey S. Steinman, Douglas R. Hardy (steinman@ramlabs.com, douglas.hardy@navy.mil)
Abstract
This article proposes the standardization of a layered simulation architecture that addresses many of the critical modeling needs of the DoD simulation community. The Standard Simulation Architecture works with HLA to provide the additional infrastructure that is necessary for developing highly interacting yet decoupled software models, while simultaneously supporting technology infusion from R&D organizations. The overall goal is to provide a cost-effective solution that features complete flexibility for simulation systems without sacrificing performance. A layered architecture is proposed to modularize critical capabilities such as high-speed communications between nodes in a multiprocessing federate, general-purpose software utilities, modeling semantics, support for human behavior representation, time management, interest management, and automated interoperability with HLA. The interface of each layer must be standardized to promote (1) model development and composability, (2) portability and interoperability with other models, (3) scalable high performance, and (4) technology infusion from the research community. The Standard Simulation Architecture extends interoperability and reuse principles taken from the High Level Architecture to (1) the entities residing within a multiprocessing federate and to (2) the components hierarchically residing within an entity or within other components. This standardized hierarchical modeling paradigm promotes the development of a reusable entity and component repository that can be reused to support different modeling applications.
HLA Principles Applied to SSA
Within the DoD simulation arena, the discussion of interoperability and reuse has centered on HLA. Four important interoperability principles have emerged from the development of HLA that are directly applied to the SSA.
1. A standardized software framework with well-defined interfaces is required to interconnect reusable models (e.g., the RTI).
2. The data exchanged by the models must follow an agreed upon standard (e.g., the FOM).
3. Distributed object technology allows models to (1) know about each other’s state and (2) invoke actions within other models in a coordinated manner (e.g., TM, OM, DM, DDM, and OWN).
4. The double-abstraction barrier principle allows a model to invoke actions on other models while hiding the details concerning which specific models are participating in the action and which methods those participating models provide to handle the action (e.g., Interactions).
HLA interoperability and reuse principles can be applied within the proposed Standard Simulation Architecture to address three distinct levels of granularity:
1. Federates within an HLA Federation
2. Entities within a parallel or sequential Federate
3. Components hierarchically composed within an entity
One of the goals of the Standard Simulation Architecture is to facilitate high-performance interoperability and reuse for both (1) technology insertion, and (2) models through the creation of entity and component repositories.
High-Level Modeling Concept
The Standard Simulation Architecture promotes high-speed interoperability and reuse at three different levels. First, Federates can interoperate through HLA interfaces using the HPC-RTI, or they can interoperate directly within the Standard Simulation Architecture. All HLA federates (including SSA Federations) interoperate through a well-defined FOM and through standard usage of the RTI. An example of eleven interoperating Federates is shown in Figure 1.
Figure 1: Interoperability between Federates in the Standard Simulation Architecture.
Second, entities within the Standard Simulation Architecture interoperate in parallel through the DSMS layer. This means that entities obtain information about other entities by subscribing to each other’s published Federation Objects. Entities process events scheduled by other entities using the formal DSMS Interaction mechanism. This not only supports the parallel processing paradigm, but also maintains the important abstraction that interacting entities could reside within different HLA Federates.
Third, components within entities interoperate through fully specified type-checked interfaces using polymorphic functions and methods. Models are hierarchically composed of components within an entity to support arbitrary levels of fidelity and detail. Like entities, components also coordinate the publication and subscription of Federation Objects and interactions with interest management. This is shown with a UML class diagram in Figure 2.
Figure 2: Entities, components, and FoMgrs. Hierarchical components are used to decompose an entity model into sub-models.
From a different perspective, another way to visualize the different levels of granularity within the architecture is to consider Inter Process Communication (IPC) mechanisms. The UML Diagram in Figure 3 shows the hierarchical decomposition of an HLA Federation as it relates to the different levels of IPC granularity.
Figure 3: A UML diagram showing the hierarchical decomposition of an HLA
Federation in the Standard Simulation Architecture.
The Standard Simulation Architecture This section provides an overview of the layered architecture. For more details see the full paper, “Evolution of the Standard Simulation Architecture”, Steinman and Hardy, 04S-SIW-100 Proceedings, April 2004.
The proposed Standard Simulation Architecture is shown in Figure 4. It is comprised of multiple software layers that simulation systems build upon.
Figure 4: The Standard Simulation Architecture.
The standardization process defines the set of interfaces for each of these layers. Once this is accomplished, different implementations of these layers can be combined to form complete simulation infrastructures that may be optimized for different types of simulations, communication networks, computing platforms, operating systems, languages, and compilers.
The System Services, Threads, Network Communications, Internal High-Speed Communications and External Distributed Communications layers provide a full-spectrum of system utilities and inter-process communication services in a standard portable manner.
The Rollback Framework, Event Management Services, and Time Management layers provide the basic infrastructure that is necessary to support discrete-event and real-time simulations executing on single or multiple CPU machines.
The Utilities, Rollback Utilities, Persistence, Standard Template Library, Standard Modeling Framework, Distributed Simulation Management Services, and External Modeling Framework layers provide the basic set of constructs and tools required for software developers to efficiently build simulation models and to directly connect them to external systems such as graphical user interfaces and hardware devices. Persistence is critical for supporting checkpoint/restart and dynamic load balancing functionality.
The SOM/FOM Translation Services, HPC-RTI Interface, and HLA Gateway layers support interoperability between Standard Simulation Architecture Federates, legacy HLA Federates, and HLA Federations.
The Component Repository and the Entity Repository provide a library of models that were designed for reuse across multiple simulation domains. Note that entities interoperate through federation objects and interactions, while components interoperate through polymorphic methods.
The complete architecture provides high-speed software reuse and interoperability between SSA federates, entities, and components. It further provides interoperability with legacy Federates and HLA Federations through the HPC-RTI Interface and HLA Gateway layers. Non-HLA external systems such as high-speed hardware or specialized graphical displays may integrate and directly interoperate with the overall system through the External Modeling Framework.
Benefits
A number of important benefits will be provided to the DoD simulation community through the standardization of these layers.
1. A common infrastructure will facilitate the development of reusable Federates, Entities, and Components.
2. The layered simulation architecture will allow simulation projects to individually combine the most efficient implementations of each layer on targeted machines to achieve the best performance.
3. Optimized sequential and parallel processing capabilities will provide efficient usage of CPU resources ranging from desktop machines to massively parallel supercomputers.
4. High-speed interoperability between new models and legacy systems will be fully supported.
5. Software models will be portable to different machines, operating systems, networks, languages, and compilers.
6. The popular business models (i.e., COTS, GOTS, and Open Source) for software development are not only supported, but also encouraged.
7. A cost effective strategy is provided to focus applied research and development efforts for new technology.
Through the formation of standards, the Standard Simulation Architecture will significantly lower the cost of developing, composing, and executing simulations. It will also focus both technology and model development software for reuse, providing synergy in the DoD simulation community.
Authors
Jeffrey S. Steinman
Dr. Jeffrey S. Steinman, Vice President and Chief Scientist at RAM Laboratories received his Ph.D. in 1988 from the University of California Los Angeles in High-Energy Particle Physics. Between 1988 and 1995, Dr. Steinman led several high-performance computing R&D activities at JPL/Caltech in support of Strategic Defense, Air Defense, Ballistic Missile Defense, and NASA space exploration missions. During this time, he developed the SPEEDES parallel simulation engine.
Dr. Steinman is currently directing the development of the WarpIV Simulation Kernel at RAM Laboratories, Inc.
Douglas R. Hardy
Mr. Hardy, Project Manager at Space and Naval Warfare Systems Center, San Diego received his Master’s in 1985 from Arizona State University in Applied Mathematics and Physics. From 1986 to the present, Mr. Hardy, has led several Modeling and Simulation R&D projects in support of the Defense Advanced Research Projects Agency (DARPA), the Office of Naval Research (ONR), and a variety of multi-sponsored Advanced Concepts Technology Demonstrations (ACTDs).
Mr. Hardy is currently directing the modernization development for the Enhanced Naval Warfare Gaming System (ENWGS) program.
|