摘要
分布式处理是数据流管理中的主流技术,SWAT算法可以有效地减少聚集操作时通信量,提高分布式系统信道的使用率。文章在SWAT算法的基础上,改变不同结点的更新周期,平衡了算法的计算速度和精确度,并且使用国际通用数据集,在斯坦福大学开发的STREAM系统上进行了测试。对于分布式系统,在数据流数据速率变化比较快的情况下,用时少、误差小。
Distributed processing is a very promising route towards data stream processing model. SWAT can reduce traffic between nodes in case of aggregation. The algorithm that is in the full utilization of the SWAT changes the update cycle of different level. Thus, it can keep the balance of algorithm speed and accuracy. Using the international universal data set, it is tested in the STREAM system developed by Stanford University. For distributed system, the time usage is short and precision is high of the algorithm when speed of data stream changes rapidly.
出处
《微电子学与计算机》
CSCD
北大核心
2006年第10期28-31,共4页
Microelectronics & Computer
关键词
数据流
分布式系统
增量聚集
Data stream, Distributed enervation, Incremental aggregation