摘要： 分布式数据流系统中，当输入数据流速发生较大波动时，会引起部分节点过载，从而影响整个系统的数据处理能力。针对这个问题，该文分析了分布式数据流系统的特点以及现有负载平衡算法的局限性，提出了一种利用多层重叠结构进行负载信息收集和负载分配的平衡算法。该算法利用虚拟树形结构进行负载信息的收集和负载分配，具有较好的扩展性能。以环形虚拟结构中保存的节点相对位置信息为依据进行负载移动，从而有效降低系统的响应时间。实验结果表明该算法具有良好的负载平衡能力和一定的应用价值。

关键词: 数据流, 数据流系统, 负载平衡, 重叠网络

Abstract: In distributed data stream processing systems, part nodes overload caused by input data stream fluctuation may reduce data processing ability of the whole system. Aiming at the practical problem, based on the characteristics of distributed data stream processing system and shortcomings of existing load-balancing methods, a multi-layer overlay based load balancing algorithm is proposed. Multi-layer overlay structure is used to collect load information of the system and redistribute load among node. Numerical experiment shows the well load-balancing ability and practicability of the algorithm.

Key words: data stream, data stream management system, load balance, overlay network

中图分类号: 

• TP393