An Approach to Preventing the Effect of Traffic Spikes and Its Performance Study

  • Yutaro Onoue
    Graduate School of Software and Information Science, Iwate Prefectural University
  • Jiahong Wang
    Graduate School of Software and Information Science, Iwate Prefectural University
  • Eiichiro Kodama
    Graduate School of Software and Information Science, Iwate Prefectural University
  • Toyoo Takata
    Graduate School of Software and Information Science, Iwate Prefectural University

Bibliographic Information

Other Title
  • データ間の関連性を利用したトラフィックスパイクの波及に対する予防手法の提案とその評価

Search this article

Description

In recent years, dynamic distributed storage services such as Amazon Dynamo and IIJ GIO have been provided, which is characterized by dynamically adding and removing servers according to users' needs to optimize system performance. For these storage systems, however, there is the possibility that system responsiveness decreases suddenly for sudden increase in access traffic. This paper gives an approach to solving the problem. In the proposed approach, the concepts of causal association and cooccurrence association among Web data have been introduced, data that will become system performance bottleneck according to the associations are predicted, replicas of the target data will be distributed to some other servers when the associated data have become hotly accessed, so that traffic loads could be balanced. We have confirmed the effectiveness of the proposed approach by experiments.

Journal

Keywords

Details 詳細情報について

  • CRID
    1574231877670848384
  • NII Article ID
    110009831684
  • NII Book ID
    AA12628305
  • Text Lang
    ja
  • Data Source
    • CiNii Articles

Report a problem

Back to top