KAF: Kalman Filter Based Adaptive Maintenance for Dependability of Composite Services

  • Authors:

  • Huipeng Guo;Jinpeng Huai;Yang Li;Ting Deng

  • Affiliations:

  • School of Computer Science and Engineering, Bei Hang University, Beijing,;School of Computer Science and Engineering, Bei Hang University, Beijing,;School of Computer Science and Engineering, Bei Hang University, Beijing,;School of Computer Science and Engineering, Bei Hang University, Beijing,

  • Venue:
  • CAiSE '08 Proceedings of the 20th international conference on Advanced Information Systems Engineering
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Service composition is fundamental in development of Web service oriented applications. Dependability of composite services is of significant importance since it directly impacts users' experience. However, dependability of a composite service may change over time as a result of inevitable changes in component services. In addition, users may also pose varying dependability requirements to meet different needs. It has become a big challenge to dynamically maintain the dependability of composite services. This paper proposes an innovative system called KAF that constructs a closed-loop control for adaptive maintenance of composite services. Modeling the control process as a Markov decision process (MDP), we further design an efficient Kalman-Filter based algorithm for service state prediction. With the availability of the precise prediction, optimal control decisions can be made. We evaluate the performance of KAF against other alternative approaches through comprehensive experiments and results demonstrate that KAF is capable for adaptive dependability maintenance.