By Topic

Dynamic load balancing for distributed network management

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$31 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

3 Author(s)
Yoshihara, K. ; KDDI R&D Labs. Inc., Saitama, Japan ; Isomura, M. ; Horiuchi, H.

The scalability limitations of centralized management models have motivated distributed management models, in which management programs describing some management tasks are distributed and executed on managed systems. In the models, management program distribution that considers dynamic network resource utilization is one of the most important challenges, in striking a load balance between management and managed systems for an entire managed network. Some methods for load balancing have been studied; however, they cannot adequately be achieved throughout an entire managed network. This arises from criteria for load balancing that lacks dynamic network resource utilization, or from a localized subnetwork in which the performance is limited, although it does include processing loads for dynamic network resource utilization. To solve this, a new dynamic load balancing method is proposed for distributed network management. Thus, systems that execute management programs are decided dynamically on the basis of CPU utilization for each system and the bandwidth required for executing all management programs. Two typical algorithms derived from the proposed method, each having different criteria in the form of mean deviation and range types with respect to CPU utilization, are introduced. They were evaluated analytically. according to capability, i.e., how well they perform as close to load balancing as possible, as well as time complexity. The results show that the mean deviation type algorithm performs better at almost the same computational cost. A prototype system is also implemented based on the proposed method, and evaluated empirically by applying it to an operational LAN. The proposed method performs well in trials with a trivial overhead.

Published in:

Integrated Network Management, 2003. IFIP/IEEE Eighth International Symposium on

Date of Conference:

24-28 March 2003