摘要
在大型分布式仿真环境下,仿真的顺利推进取决于每一个邦员的健康运行,负载不平衡会对仿真性能造成严重的影响。作为分布式仿真事实上的标准,HLA仿真框架主要关注于解决分布式仿真的互操作和可重用问题,而没有提供对可伸缩性、动态负载平衡以及容错的支持。通过分析HLA仿真框架的不足,提出了一个基于网格服务的先进分布式仿真框架GADS,并在此基础上设计了一个负载平衡解决方案,随后重点描述了其中的负载平衡策略和基于仿真代理的高效邦员迁移算法。利用在GADS框架中引入的仿真代理层,该迁移方案大大降低了迁移开销,并实现了真正的"免冻结"迁移。
As the IEEE 1516 standard for distribution simulation,High Level Architecture(HLA) framework does not solve the issues of scalability,dynamic load-balance and fault tolerance.HLA does not support federation migration,which would balance the system workload on heterogeneous distributed resources and consequently improve simulation’s performance.In this paper,a Grid based Advanced Distributed Simulation(GADS) framework is proposed.Based on the GADS framework, this paper proposes a highly efficient load balance strategy and a federate migration protocol by utilizing grid computing.A simulation broker layer is introduced which is in charge of the federates’communication with the other simulation federates. Experiment shows that this approach decreases the migration overhead,improves the performance of the simulations in large-scale distributed environments,and realizes the"freeze-free"migration.
出处
《计算机工程与应用》
CSCD
北大核心
2011年第9期25-28,36,共5页
Computer Engineering and Applications
基金
国家部委2009年度预研基金
关键词
高级体系结构(HLA)
分布式仿真
邦员迁移
仿真代理
网格计算
High Level Architecture (HLA)
distributed simulation
federate migration
simulation broker
grid computing