【24h】

A Fault-Tolerant Routing Strategy with Graceful Performance Degradation for Fat-Tree Topology Supercomputer

机译:胖树拓扑超级计算机容忍性能下降的容错路由策略

获取原文

摘要

In recent years, in order to solve the problem of high-speed signal transmission quality, the solution of short reach HSS + on-board optics + passive optical fiber is gradually replacing the original long reach HSS + AOC inter-switch interconnection scheme. When the on-board optics fails, its replacement time will be much higher than that of AOC, which increases the duration of interconnection failure in the system. We believe that it is essential that the fault-tolerant routing strategy can reduce the degradation of the performance of the interconnection network without suspending the operation of the tasks, as this will greatly improve the overall availability of the system during the duration of the failure. In this paper, a fault-tolerant routing strategy for fat-tree topology is proposed, which can tolerate multiple interconnect failures with only a graceful performance degradation through actual system testing.
机译:近年来,为了解决高速信号传输质量问题,短距离HSS +板载光学+无源光纤的解决方案逐渐取代了原来的长距离HSS + AOC交换机间互连方案。当板载光学器件发生故障时,其更换时间将比AOC的更换时间长得多,这会增加系统中互连故障的持续时间。我们认为至关重要的是,容错路由策略可以在不暂停任务运行的情况下减少互连网络性能的下降,因为这将在故障期间极大地提高系统的整体可用性。本文提出了一种胖树拓扑的容错路由策略,通过实际的系统测试,该策略可以容忍多个互连故障,而性能只会出现适度的下降。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号