基于AWGR的光电混合交换高性能计算互连网络  被引量:2

An optical-electrical hybrid switching high performance computing interconnect network based on AWGR

在线阅读下载全文

作  者:王昭[1] 裴秉玺 何卫锋[1] WANG Zhao;PEI Bing-xi;HE Wei-feng(Department of Micro-Nano Electronics,Shanghai Jiao Tong University,Shanghai 200240,China)

机构地区:[1]上海交通大学微纳电子学系

出  处:《微电子学与计算机》2019年第8期19-24,共6页Microelectronics & Computer

摘  要:为了满足高性能计算机中互连网络的高吞吐量、低延迟需求,本文基于阵列波导光栅路由器(AWGR),提出了一种低延迟的光电混合交换的互连网络结构.该网络以"天河二号"超级计算机的互连网络拓扑结构为基础,使用无竞争阻塞的AWGR替代了部分电路由器,通过改进设计电路由器的光端口来实现与原有高阶路由器内部的电交换方式兼容.在此基础上,使用OMNeT++软件框架编写了新的光电混合交换网络的仿真模型.仿真结果表明,与原始网络相比,光电混合交换网络的总延迟降低了8%~10%.To satisfy the throughput and latency requirement of high-performance-computing interconnect network, a low-latency optical-electrical hybrid switching network structure based on arrayed waveguide grating router(AWGR) is proposed. On the basis of "Tianhe-Ⅱ" supercomputer’s network topology, part of electrical high-radix routers are replaced by non-blocking and contention-free AWGRs. The optical port of electrical router is re-designed to interface the optical packets with router’s internal switching logics. To evaluate the performance of proposed network, a simulation model is developed using OMNeT++ software framework. The simulation shows that the data latency of hybrid switching network is 8%~10% lower than the electrical switching network.

关 键 词:高性能互连网络 阵列波导光栅路由器 光分组交换 网络延迟 

分 类 号:TP391.9[自动化与计算机技术—计算机应用技术]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象