Advertisement

A Survey of Dynamic Load Balancing

  • Hisao Kameda
  • Jie Li
  • Chonggun Kim
  • Yongbing Zhang
Part of the Telecommunication Networks and Computer Systems book series (TNCS)

Abstract

Static policies use only the system statistical information in making load balancing decisions, and therefore have the advantages in mathematical analysis and in implementation because of their simplicity. They do not, however, adapt to fluctuations of workload. Under a situation where the system workload is statistically balanced, some computers may be heavily loaded at a given instant (hence suffering from performance degradation), while others are idle or lightly loaded. On the other hand, dynamic policies [ELZ86a, WM85, Zho88, GDI93, MTS90, SK90] attempt to balance the workload dynamically corresponding to the current system state and are therefore thought to be able to further improve system performance. Dynamic policies are, however, much more complex than static policies. Studies concerning on dynamic load balancing have been usually limited to specific models that assume either that all the nodes in the system are identical or that overheads involved in load balancing are negligible [MTS89, MTS90, ELZ86a, ELZ86b, LM82, VD90]. In order to take account of the current system state in load balancing decisions, much information is needed, e.g., load state of each node in the system, information on each job, and the congestion information of the network. The schemes for information collection, job selection for transfer, and job allocation are also important and affect the performance of a load balancing policy. In this chapter the performance metrics for dynamic load balancing, the key components of a load balancing policy, and a brief review of dynamic load balancing policies are described.

Keywords

Load Balance Queue Length Load Information Transfer Policy Dynamic Load Balance 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

Copyright information

© Springer-Verlag London Limited 1997

Authors and Affiliations

  • Hisao Kameda
    • 1
  • Jie Li
    • 1
  • Chonggun Kim
    • 2
  • Yongbing Zhang
    • 3
  1. 1.Institute of Information Sciences and ElectronicsUniversity of TsukubaTsukuba-shi, IbarakiJapan
  2. 2.Department of Computer EngineeringYeungnam UniversityGyongsan, KyongbukKorea
  3. 3.Institute of Policy and Planning SciencesUniversity of TsukubaTsukuba-shi, IbarakiJapan

Personalised recommendations