Research On Key Technologies Of Service-Oriented Semantically Composable Simulatio

Composable simulation is one key technology unsolved within the new extensible web-based modeling and simulation framework, i.e. XMSF (Extensible Modeling and Simulation Framework). Especially semantic composability problem, a crux in composable simulation, has aroused great attention within M&S (Modeling and Simulation). Semantically composable simulation promises to solve rapid and efficient simulation construction across WAN (Wide Area Network), based on which web-based simulation application will no longer only serve for human, but for the computer program as well. Under this ideal circumstance, the interaction and integration of simulation applications will be automatically achieved. However, it is a long way ahead and a large amount of work left unresolved before this goal can be achieved, therefore based on development requirements of current distributed simulation, combined with technologies of Web services and Semantic Web, we choose several key problems in semantically composable simulation to start our research.First of all, the study of conceptual framework of semantically composable simulation is carried out. Relative concepts, complexity characteristic and hiberarchy of semantically composable simulation are detailedly analysed, and the concept of simulation semantic composability is explicitly defined. Further, the feasibility and necessity of Web services, SOA (Service-Oriented Architecture) and Semantic Web becoming the supporting technologies of semantically composable simulation are demonstrated. Aimed at M&S's requirement, a particular Web service, i.e. simulation service, is defined, and differences of simulation services from general Web services are analysed. With the above research, we establish the idea of research on service-oriented semantically composable simulation, and present a concept framework for semantically composable simulation, where a three-layered architecture including user view, component view and application view is set up. We illuminate functions provided by each layer, and determine key technologies needed to be solved in the concept framework.Building semantic description model for simulation service and realizing semantic discovery of simulation service are bases of semantically composable simulation. Based on WSDL/OWL-S we build files of SS-WSDL, SSProfile, SSModel and SSGrounding as simulation service's semantical description model, which describe simulation service from abstract semantic information to concrete deployment realization. Aimed for semantic discovery, we put forward the concept of simulation domain ontology, based on which to formalize simulation service and simulation requirement. Under these definitions we present a simulation domain ontology-based distributed simulation services semantic publication and discovery method. This method is realized to publish simulation services belonging to one same simulation domain ontology to one same registry node, and according to concept subsumption relationship between those ontologies distributed registry nodes are connected as a directed connected tree. With the tree structure we realize a request messages transmit mechanism, which can efficiently reduce the amount of matching computation and message transmitting during service discovery.As for the realization of simulation service matching during semantic discovery, we propose an ontology-based multiple hierarchies matching algorithm for simulation services (OMHM4SS), in which we construct a concept category tree of simulation domain ontologies, partition simulation services according to domain concept, and realize to choose the domain area for simulation requirement.OMHM4SS has two layers:domain matching at concept layer and service capability matching at service layer, and also correlates with the process of service publication and service discovery: during publication simulation services are published to the registry node with the ontology they belong to; during discovery simulation requirements can be directly sent to the registry node with concepts relative to their domain ontologies.QoS (Quality of Service) has to be considered during simulation services composition across WAN. Due to the dynamic characteristic and uncertainty of WAN, simulation services that satisfy simulation requirements in functional attributes may not always satisfy system requirement in non-functional attributes, i.e. QoS. Therefore we further study QoS-driven simulation services composition (Q-SSC) problem after semantic discovery. With deep analysis of characteristics of distributed simulation tasks, we define QoS model for composite simulation service, propose a computing and measuring standard for the model, and model Q-SSC problem as a typical multi-objective optimization problem. According to simulation users'different preferences, we solve Q-SSC problem in two different ways, and provide two different algorithms:SOOCH and MGAOOC. SOOCH assign weight vector to QoS attributes and add them together to achieve one single objective function. Using convex hull as heuristic information to solve constrainted single objective optimal problem, SOOCH has high efficiency. The shortage of SOOCH is that it needs simulation users to provide weight for each QoS, which requires users to have relative knowledge about the problem. And under the specific weight, SOOCH can only provide single solution for the user, which leaves the user no other choice. On the other hand, MGAOOC is based on multi-objective genetic algorithm, and realized to optimize multiple QoS attributes at the same time. MGAOOC can achieve a group of Pareto optimal solutions to satisfy simulation users. Alghough MGAOOC has worse efficiency than SOOCH does, it can provide more options for the user, which demonstrate that it has more flexibility and agility.After the solution for simulation services composition has been obtained, we need to monitor the execution of composite simulation system, and make sure the system running in a more efficient way. We present a method of dynamic simulation service composition, which includes a complete recomposition algorithm (SOOCH-complete, MGAOOC-complete) and a partial recomposition algorithm (SOOCH-partial, MGAOOC-partial). This method supports the user to replaning service composition. Service fault-tolerance is the way to improve simulation service's availability. We deeply analyse the environment where composite simulation service runs, dispose composite simulation services fault-tolerant problem as a network failure information detection problem, and put forward a partial synchronization model based network failure detection algorithm.In the end, based on the above theories, we design and realize service-oriented semantically composable simulation prototype system (SOSCSim), in which simulation service semantic discovery module, simulation service optimal composition module and simulation service fault-tolerant module have been simply developed. Under Eclipse IDE, with Globus toolkit we realize a simulation server and a simulation client in SOSCSim, and use a simple simulation application to partly validate the prototype system.