光互连网络中资源预留的多级光互连仲裁机制
发布时间：2018-12-15 20:18
【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%。
[Abstract]:A multilevel optical interconnection arbitration mechanism based on resource reservation strategy is proposed to achieve fast and efficient arbitration by grading the network. The transport node design of multi-priority data cache queue provides different traffic differential transmission, and realizes the network completely fair and 100% high throughput through the reservation two-level arbitration mechanism. The reasonable layout of the fast arbitration channel is designed and the arbitration delay is greatly shortened. The simulation results show that the hierarchical arbitration strategy based on resource reservation can provide fair service to all nodes in various traffic modes. Compared with Feather Weight, the throughput of hierarchical arbitration strategy is increased by 17. Compared with 2-pass, arbitration delay is reduced by 15 and power consumption is reduced by 5.
【作者单位】： 国防科技大学计算机学院;
【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)
【分类号】：TN491
，
本文编号：2381228
[Abstract]:A multilevel optical interconnection arbitration mechanism based on resource reservation strategy is proposed to achieve fast and efficient arbitration by grading the network. The transport node design of multi-priority data cache queue provides different traffic differential transmission, and realizes the network completely fair and 100% high throughput through the reservation two-level arbitration mechanism. The reasonable layout of the fast arbitration channel is designed and the arbitration delay is greatly shortened. The simulation results show that the hierarchical arbitration strategy based on resource reservation can provide fair service to all nodes in various traffic modes. Compared with Feather Weight, the throughput of hierarchical arbitration strategy is increased by 17. Compared with 2-pass, arbitration delay is reduced by 15 and power consumption is reduced by 5.
【作者单位】： 国防科技大学计算机学院;
【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)
【分类号】：TN491
，
本文编号：2381228
本文链接：https://www.wllwen.com/kejilunwen/dianzigongchenglunwen/2381228.html
