Overhead and Performance of Dynamic Network Slice Allocation for Mobile Users

Abstract

Based on virtualisation technologies like SDN and NFV, modern networks like 5G create customised logical networks known as Network Slicing to serve distinct user groups. The mobility of a significant part of these users brings great challenges for that scenario, e.g. demand variations and continuous connectivity. In such a scenario, the dynamic resource allocation of these slices is shown as a possible solution. However, that approach may be sensitive to computing overhead and trade-off scenarios, which are not fully evaluated in the literature. This work focuses on that gap by researching the impact of network slice reconfiguration overhead on the overall network performance. An extensive evaluation of Dynamic Network Slicing in providing user mobility support in Fog computing-based architectures is presented. This paper presents a resource allocation model that calculates the required time to allocate a slice instance. Furthermore, based on those values, this work presents an evaluation of the impact of resource allocation delays in terms of bandwidth allocation, latency, and service outage in different dynamic slicing scenarios that include service migration. Simulations based on realistic values point out that reducing the allocation delay from 60 to 30 s can lead the network to 25% more stable latency to its users in 60 min slice lifetime.