摘要
构建一种微细粒度故障模型,有效地区分链路故障和通道故障,并在该模型的基础上提出一种基于故障节点再利用的容错路由算法.本算法由一个动态暂存机制(dynamic pseudo-receiving mechanism,DPR)和动态转向控制(dynamic turn management,DTM)构成,分别用于暂存数据包和避免死锁.最后实验结果表明,本文算法较先前的容错算法,时间延迟降低27%,吞吐量提高21%.尤其在数据包注入率增大时,本文方案有着明显的优势。
This paper put forward a fine-grained fault-tolerant model which can effectively distinguish between the link failure and the channel failure,and a fault-tolerant routing algorithm was proposed with the model based on the reuse of fault nodes. The algorithm consists of a dynamic pseudo-receiving mechanism (DPR) and a dynamic turn management(DTM) , which are respectively used for temporarily storing data packets and avoiding the deadlock. The experimental results show that the method can reduce the delay at most by 27% and increase the throughput by 21% on the basis of less hardware overhead compared with the previous fault-tolerant routing algorithm. When the injection rate of the data packets is increased, the method has the obvious advantages, ensuring the reliability of the network.
出处
《小型微型计算机系统》
CSCD
北大核心
2017年第5期997-1001,共5页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61572520)资助
关键词
细粒度故障模型
通道故障
动态暂存机制
动态转向控制
fine-grained failure model
channel fault
dynamic pseudo-receiving mechanism
dynamic turn management