计算机应用 ›› 2010, Vol. 30 ›› Issue (3): 736-739.

• 网络与通信 • 上一篇    下一篇

改进型AR(1)模型在域名系统中的应用

郑伟1,谷大武2,陆海宁2   

  1. 1. 上海交通大学计算机系
    2. 上海交通大学信息安全学院
  • 收稿日期:2009-09-14 修回日期:2009-10-27 发布日期:2010-03-14 出版日期:2010-03-01
  • 通讯作者: 郑伟

Application of improved AR(1) model in DNS

  • Received:2009-09-14 Revised:2009-10-27 Online:2010-03-14 Published:2010-03-01
  • Contact: Wei Zheng

摘要: 服务器选择算法是域名系统(DNS)服务器处理迭代查询时的核心算法。在DNS中的所有查询请求中,迭代查询所占的比例大于30%,因此服务器选择算法的性能对整个DNS服务器的整体性能有着非常大的影响。简要分析了现有的服务器选择算法,指出其优点和缺点。同时,基于已有的AR(1)模型,提出了一种改进型AR(1)自回归模型,该模型利用历次服务器响应时间构成的时间序列,采用动态预测的方法来预测服务器响应时间。该模型能够有效避免由于网络拥塞状况造成的性能波动和服务器短时故障造成的性能损失,大幅提高了服务器选择算法的性能。同时,经过改进的AR(1)模型的适用范围得到扩大,可以适用于所有DNS。

关键词: 域名系统, AR(1), 时间序列预测, 服务器选择算法, 迭代查询, Unbound

Abstract: Server selection algorithm is the key algorithm for Domain Name System (DNS) while handling iterative queries. Among all queries sent to DNS, the proportion of iterative query is larger than 30%, so the performance of server selection algorithm directly affects the performance of a DNS server. The existing server selection algorithms were briefly reviewed and both advantages and disadvantages of these algorithms were described. Then, an improved AR(1) auto-regressive model was proposed. Through this new model, the response time of DNS server could be dynamically predicted by using previous response time series. This new model can efficiently avoid performance fluctuation and loss due to network congestion and short-time system failure. At the same time, the application scope of AR(1) model is broadened by the new model and it is suitable for all DNS.

Key words: DNS(Domain Name System), AR(1), time series prediction, server selection algorithm, iterative query, Unbound