近年研究发现网络中的业务量呈自相似特征,这种自相似特征显著影响网络的流量控制与排队性能,本文在自相似网络流量可预测的基础上,利用线性回归分析理论进行流量预测,并应用控制理论中的预测PI控制器原理设计了动态矩阵PI控制主动队列...近年研究发现网络中的业务量呈自相似特征,这种自相似特征显著影响网络的流量控制与排队性能,本文在自相似网络流量可预测的基础上,利用线性回归分析理论进行流量预测,并应用控制理论中的预测PI控制器原理设计了动态矩阵PI控制主动队列管理(Dynam icMatrix PI Control-Active QueueManagement,简称DM-PIC-AQM)算法,以克服队列的剧烈振荡,保持队列稳定在期望的长度.仿真实验结果表明,DMPIC-AQM算法在网络流量剧烈变化和小期望队列长度的情形下,DMPIC-AQM算法明显优于RED与PI算法.展开更多
Network management has widely adopted multi-agent architecture, and AGIMA is one of the successful prototypes that have developed mobile code facilities and group management functions. Taking consideration on the requ...Network management has widely adopted multi-agent architecture, and AGIMA is one of the successful prototypes that have developed mobile code facilities and group management functions. Taking consideration on the requirements of exchange and sharing of network management information between agents, the network management information described by next generation structure of management information can be represented as knowledge by XML-based resource description framework. These efforts need integration of some newer toolkit software and become foundation for introduction of network management intelligence into agents.展开更多
文摘近年研究发现网络中的业务量呈自相似特征,这种自相似特征显著影响网络的流量控制与排队性能,本文在自相似网络流量可预测的基础上,利用线性回归分析理论进行流量预测,并应用控制理论中的预测PI控制器原理设计了动态矩阵PI控制主动队列管理(Dynam icMatrix PI Control-Active QueueManagement,简称DM-PIC-AQM)算法,以克服队列的剧烈振荡,保持队列稳定在期望的长度.仿真实验结果表明,DMPIC-AQM算法在网络流量剧烈变化和小期望队列长度的情形下,DMPIC-AQM算法明显优于RED与PI算法.
文摘Network management has widely adopted multi-agent architecture, and AGIMA is one of the successful prototypes that have developed mobile code facilities and group management functions. Taking consideration on the requirements of exchange and sharing of network management information between agents, the network management information described by next generation structure of management information can be represented as knowledge by XML-based resource description framework. These efforts need integration of some newer toolkit software and become foundation for introduction of network management intelligence into agents.