Hierarchical network management: a scalable and dynamic mobile agent-based approach

  • Authors:

  • Damianos Gavalas;Dominic Greenwood;Mohammed Ghanbari;Mike O'Mahony

  • Affiliations:

  • Dimokratias 54, Neo Psychiko, 15451, Athens, Greece and Communication Networks Research Group, Electronic Systems Engineering Department, University of Essex, Wivenhoe Park, Colchester, UK;Network Agent Research, Fujitsu Laboratories of America, Inc., 595 Lawrence Expressway, Sunnyvale, CA;Communication Networks Research Group, Electronic Systems Engineering Department, University of Essex, Wivenhoe Park, Colchester, UK;Communication Networks Research Group, Electronic Systems Engineering Department, University of Essex, Wivenhoe Park, Colchester, UK

  • Venue:
  • Computer Networks: The International Journal of Computer and Telecommunications Networking
  • Year:
  • 2002

Quantified Score

Hi-index 0.00

Visualization

Abstract

Several distributed management architectures, incorporating mobile agent (MA) technology, have been recently proposed to answer the scalability limitations of centralised models and the flexibility problems of static hierarchical frameworks. Yet, although agent-based management frameworks have recently started evolving from the early 'flat' models to hierarchical structures, they cannot efficiently cope with the dynamically changing traffic and topological characteristics of modern networks. This is mainly due to the limited use of agent mobility (employed either through mid-level manager entities or between static mid-level managers and managed devices) and lack of appropriate policies enabling automatic calibration of the management system based on network conditions. This paper presents a hierarchical agent-based infrastructure, suitable for the management of large-scale enterprise networks that addresses these issues. The transition to hierarchical agent-based management is achieved through a mid-level manager that being a MA itself, operates at an intermediary level between the manager and the legacy systems and takes full control of managing a given network segment. These entities make the system more adaptive to changing networking conditions, while localising the traffic associated with bandwidth-intensive monitoring applications. A quantitative evaluation, in terms of the overall management cost, confirms that this architecture outperforms both centralised approaches and MA-based 'flat' management models.