Dynamic reallocation of capacity in logically fully-connected networks

We study the benefits that can be realised when capacity reallocation schemes are implemented in telecommunications networks. Our basic model consists of a (possibly sparse) physical network above which is overlaid a logically fully‐connected network. This is achieved by reserving capacity on the physical links for exclusive use by each origin–destination pair. If the capacity allocated to the origin–destination pairs is fixed, then such a network is likely to produce poor performance due to restriction of multiplexing opportunities. However, this effect can be largely negated by employing a capacity reallocation scheme, where capacity is moved in a dynamic fashion to areas where congestion is occurring. For such a scheme to be scalable, it is essential that it be distributed in nature. In this paper, we present three distributed, dynamic capacity reallocation schemes and compare their performance using a simulation model. Copyright © 2004 John Wiley & Sons, Ltd.