Skip to Main Content
Resource allocation is a fundamental problem in connection-oriented networks. A preemption mechanism would provide available and reliable services to new connections with higher priority by tearing down existing connections of lower priority. However, end users of the existing connections that are preempted will suffer service disruptions. The duration spent and the work done for these connections are wasted, leading to lower useful utilization of the overall network resources. Soft preemption can alleviate this by rerouting a connection that is about to be preempted before actually tearing it down so that the interruption of ongoing service can be avoided. In this paper, we focus on minimizing the service disruptions caused as a result of preemption by proposing algorithms that incorporate the soft preemption feature. A centralized algorithm is developed to select the network links that have a high number of reroutable connections in order to minimize service disruptions. For feasible deployment, a decentralized preemption algorithm that uses local information is subsequently proposed. Simulation results indicate that these approaches not only reduce the service disruption but also lead to higher network throughput than what can be achieved by existing preemption algorithms.