Reconfiguring Network Slices in 5G Networks

5G network slices independently scale in and out their Virtual Network Functions (VNFs) according to the traffic load during the day. Nevertheless, the lack of coordination between these adaptations leads to undesired conditions, such as wasted resources and congested links, which increase the risk of Service-Level Objectives (SLOs) violations. Network operators can thus benefit from the ability to reconfigure optimally network slices multiple times per day, but the stringent requirements of 5G use cases make reconfiguration a particularly hard and time-consuming task. In this paper, we provide two key contributions to the study of network slices reconfiguration. We describe the first dataset on network slices reconfiguration in 5G networks, and we show that state-of-the-art techniques require excessive time to reconfigure network slices. We then present Matryoshka, a divide-and-conquer approach that can scale with the number and complexity of network slices by leveraging the multiple cores available in the system. In our evaluation, we show that our approach is indeed able to reconfigure network slices in a timely fashion.