On-line routing and bandwidth allocation for elastic traffic and for its restoration

In infocommunications networks the amount of free network resources varies in time significantly. Furthermore, after a network failure a large amount of backup paths are built up and later torn down. To achieve the highest possible network throughput and availability while keeping the loss of data low and guaranteeing fairness, the bandwidth (rate) of sources should be adjusted accordingly in an elastic way. After a complex network failure the path of the demand should be changed (re-routed) as well as its bandwidth likely decreased. Existing papers study the case when the traffic matrix is given in advance and centralised management is assumed. In contrast to these we investigate the case of on-line (i.e., dynamic) routing when unpredictable backup or new demands arrive in an unforeseen sequence. For this purpose instead of the centralised management we use distributed control. We propose flexible methods that determine the bandwidth of elastic sources and the routes used by these demands simultaneously. We evaluate the algorithms according to the obtained blocking ratios, network utilisation and computational time. We show that they yield lower blocking ratio and higher network utilisation than conventional shortest path methods.