基于Heron平台的自适应重传策略  

Automatically adaptive retransmission strategy based on Heron platform

作　　者：夏毅 于炯[2] 鲁亮 张译天 李梓杨 XIA Yi;YU Jiong;LU Liang;ZHANG Yi-tian;LI Zi-yang(School of Software,Xinjiang University,Urumqi 830008,China;School of Information Science and Engineering,Xinjiang University,Urumqi 830046,China;School of Computer Science and Technology,Civil Aviation University of China,Tianjin 300300,China)

机构地区：[1]新疆大学软件学院,新疆乌鲁木齐830008 [2]新疆大学信息科学与工程学院,新疆乌鲁木齐830046 [3]中国民航大学计算机科学与技术学院,天津300300

出　　处：《东北师大学报（自然科学版）》2020年第1期55-62,共8页Journal of Northeast Normal University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61862060,61462079,61562086,61562078).

摘　　要：针对Heron容错机制中拓扑使用固定超时阈值未考虑数据源与组件内部逻辑的差异对元组往返时间影响的问题,提出了Heron环境下自适应重传策略,包括元组标记算法和最优参数估计算法.首先通过元组标记算法监控拓扑中元组,然后根据数据流的处理时延计算最优超时阈值并更新到拓扑中,从而适应当前集群的实际负载,提升集群性能.实验结果表明,自适应重传策略与默认重传策略相比,集群的元组失败率和CPU占用率在不同情况下平均降低了16%和0.6 GHz,集群吞吐量平均提高了13%.Aiming at the problem that the topology uses the fixed timeout threshold in the Heron fault tolerance mechanism without considering the difference between the data source and the internal logic of the component,this paper proposed an adaptive retransmission strategy in Heron environment,including tuple labeling algorithm and optimal parameter estimation algorithm.Firstly,the tuple was monitored by the tuple tagging algorithm,and then the optimal timeout threshold was calculated according to the processing delay of the data stream and updated to the topology,so as to adapt to the actual load of the current cluster and improve the cluster performance.The experimental results show that compared with the default retransmission strategy,the tuple failure rate and CPU usage of the adaptive retransmission strategy are reduced by 16%and 0.6 GHz on average,and the cluster throughput increased by an average of about 13%under different conditions.

关 键 词：Apache Heron 流式计算 容错 大数据 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]