光互连网络中资源预留的多级光互连仲裁机制  

Multi-level arbitration in optical network-on-chip based on resource reservation

作　　者：翦杰 赖明澈[1] 肖立权[1] JIAN Jie;LAI Mingche;XIAO Liquan(College of Computer, National University of Defense Technology, Changsha 410073 , China)

机构地区：[1]国防科技大学计算机学院,湖南长沙410073

出　　处：《国防科技大学学报》2016年第5期39-44,共6页Journal of National University of Defense Technology

基　　金：国家863计划资助项目(2015AA015302);国家重点基础研究发展计划资助项目(2012CB933504);国家自然科学基金资助项目(61572509)

摘　　要：基于资源预留策略提出一种多级光互连仲裁机制,通过将网络分级实现快速、高效的仲裁。多优先级数据缓存队列的传输节点设计,提供了不同类型流量的差异化传输;通过预约式两级仲裁机制,实现网络的完全公平与100%的高吞吐率。设计并对快速仲裁通道进行了合理布局,极大地缩短了仲裁延迟。仿真结果表明:采用基于资源预留的分级仲裁策略,在多种流量模式下所有节点均获得公平的服务。与Feather Weight相比,分级仲裁策略吞吐率提高17%;与2-pass相比,仲裁延迟减少15%,同时,功耗减少5%。A multi-level arbitration based on resource reservation in optical network-on-chip was proposed. The fast and high efficiency arbitration scheme divided the network into multi-stages. W ith the design of multi priority queues in every node, arbiters provided differential transmissions for different kinds of flows. The arbiter also presented the transmit bund resource reservation scheme to fairly reserve time slots for all nodes, thus achieving the throughput of 100% . The fast arbitration channels were proposed and designed to degrade the arbitration period, and the packet transmitting delay was reduced. The simulation results show that, with the multi-level arbitration based on resource reservation, all nodes are allocated with almost equal service under various patterns. This scheme improves throughput by 17% when compared with FeatherWeight under the self-similar traffic pattern, and eliminats arbitration delay by 15% to 2-pass arbitration, incurring total power overhead by 5 % .

关 键 词：多级光互连 快速仲裁通道 资源预留 

分 类 号：TN95[电子电信—信号与信息处理]