Abstract
As user load on web servers becomes more globalised, duplicated regional mirroring is being seen as an increasingly expensive solution to meeting regional peak demand. Alternative solutions are being explored by looking at dynamic load balancing using distributed intelligent middleware to re-route traffic from busy regions to quieter ones as global load patterns change over a 24 h cycle. The techniques used can also be employed under fault and planned maintenance conditions. One such solution, providing the load balancing via reconfigurable dynamic proxy servers, is seen as `unobtrusive' in that it works with standard web browsers and web server technology. The technique employs an evolutionary algorithm to perform combinatorial optimisation against a dynamic performance predicting model. This paper describes this solution, focussing on issues such as algorithm tuning, scalabilty and reliability. A prototype system is currently being trialled within the Systems Integration Department at British Telecommunications Research Labs, Adastral Park and is the subject of several BT held patents.
Similar content being viewed by others
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Oates, M., Corne, D. Global web server load balancing using evolutionary computational techniques. Soft Computing 5, 297–312 (2001). https://doi.org/10.1007/s005000100103
Issue Date:
DOI: https://doi.org/10.1007/s005000100103