Research And Implementation On The Highly-scalable Distributed Interactive Simulation Support Platform

Distributed interactive simulation technology has developed for decades andapplied to diverse industries. It is now enhanced with the development of computerhardware and network technology, and also modeling technology, virtual realitytechnology, software technology, and so on. On the other hand, people are bringingforward more and more requirements on distributed interactive simulation applications.Supports of distributed simulation systems are required on applications that involves inmassive complex simulation environment and large quantity of simulation nodes andentity objects, such as military simulation, online games. These requirements ask forhigh scalability of simulation support platform. However, current distributedinteractive simulation platforms have the following constraints: their systemarchitecture are not scalable so that they are unable to support join of large quantity ofsimulation nodes and dynamic objects; they allocate resources statically so that theycan not balance load, nor can they be conducive to effective utilization of resources,and for sudden collapse of certain resources in case of emergencies, they are lack ofcontingency plans; they has lower data transmission efficiency and inefficientoptimization of data filtering; in simulation applications, simulation entity objectmanagement and rendering scenes object management are separated and lack ofsupporting diversification rendering of simulation directly. All these factors have led tothe absence of high scalability in distributed simulation system.Based on the definitions and characteristics of distributed system and distributedinteractive simulation, this paper revises the previous definition of distributedinteractive simulation platform scalability, and led in-depth studies on key technologiesfor highly scalable distributed interactive simulation platform, including platformarchitecture, resource scheduling strategy, resources fault detection and federatemigration technology, efficient data transmission strategy and data filtering mechanism,integration of simulation platform and rendering platform, etc. The above researchresults are realized on HIVE, a support platform for distributed interactive simulation.Several experimental applications have been developed finally. In this paper, there areinnovations from the following four aspects:1. Definition and research methods of current distributed system and distributedinteractive simulation system follow those of traditional parallel computing system.While this paper points out it is one-sided to measure the scalability just from thenumber of processors and performance of a system, because the resources of distributed systems are heterogeneous and they can be used in diverse applications. Inthis paper, scalability of distributed interactive simulation platform is defined as theability of scaling (including scale-up and scale-down) in functionality and performanceof systems which is run on the platform when number of processors and otherhardware resources such as network change, i.e. when problem size is scaled up (orscaled down) and system resources is increased (or decreased), the performanceremains fixed and it is easy to expand (contract) the functions of simulationapplications. Compared with the traditional definition, system resources include notonly processor resources but also network resources, and system scalability includesnot only performance scalability such as CPU usage, network and accommodation forsimulation nodes but also functionality scalability such as interaction and rendering.2. This paper proposed a distributed interactive simulation platform architecturebased on three-tied model. It is composed of a global tier, group tier, and client tier, thefirst two constitute simulation compute node and the last one constitutes simulationapplication node. The three-tied model avoids bottlenecks on server side that happensin centralized server architecture while at the same avoid shortcomings such asinefficient management and operation that happens in peer-peer systems, and has betteraccommodation for simulation nodes in organization structure. In simulation resourcemanagement of the architecture, for compute nodes, the adoption of server clustertechnology based resource scheduling system design balances load well on serverclusters by separation of tasks and task allocation algorithms, and the adoption ofheartbeat technology as resource availability monitoring augments resourcemanagement ability and highlights system availabiity for large-scale simulationapplication systems. Federate migration technology in application nodes that is studiedon the basis. The architecture raises the overall efficiency and the overall capacity ofthe simulation platform, thus achieving better scalability for simulation platforms.3. This paper proposed an efficient communication strategy and data filteringmechanism. The communication strategy improves transmission efficiency byselecting different transmission mode according to various types of data exchanged indistributed interactive simulation systems. The data filtering mechanism also reduceunnecessary data transmission by adopting global-local two-layer data managementand three-layer data filtering mechanism according to features of simulation data. Inthis way, it reduces the demand of data transmission bandwidth, increases networkcapability and enhances transmission efficiency, and increases the simulation platformscalability.4. Combining research and analysis on HLA simulation application objects organization and management, this paper proposes the Unified Object Model whichconsists of Heterogenous Scene Graph tree, Action List and Universal Access Interface,realize unified management and efficient organization of scenes objects and simulationentity objects, establishes an efficient data exchange bridge between HLA simulationplatform and rendering platform, reduce the workload of integration between the twoplatform, and realizes supporting diversification rendering of simulation. It supportsnot only the traditional stand-alone rendering, also supports parallel rendering, thusimproves the functionality scalability of simulation platform.