Many load balancing algorithms balance the load according to a certain topology. Its choice can significantly influence the performance of the algorithm. We consider a two phase balancing model. The first phase calculates a balancing flow with respect to a topology by applying a diffusion scheme. The second phase migrates the load according to the balancing flow. The cost functions of the phases depend on various properties of the topology; for the first phase these are the maximum node degree and the number of eigenvalues of the network topology, for the second phase these are a small flow volume and a small diameter of the topology. We compare and propose various network topologies with respect to these properties. Experiments on a Cray T3E and on a cluster of PCs confirm our cost functions for both balancing phases.
CITATION STYLE
Decker, T., Monien, B., & Preis, R. (2000). Towards optimal load balancing topologies. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 1900, pp. 277–287). Springer Verlag. https://doi.org/10.1007/3-540-44520-x_37
Mendeley helps you to discover research relevant for your work.