2023 13th International Workshop on Resilient Networks Design and Modeling (RNDM)
2023
To provide tailored services for diverse use cases, 5G networks will use network slicing.In network slicing, multiple "slices" corresponding to each use case are hosted on a shared physical infrastructure.In the network function virtualization (NFV) paradigm, these slices are virtualised and the service is provided by routing and processing traffic through an ordered sequence of virtual network functions (VNF).The quality of the service (QoS) depends on the quantity and relative placement of the VNFs, and is quantified by a set of key performance indicators (KPIs), in a service level agreement (SLA): a contract reached between the internet service provider (ISP) and customer.In order to slice the network in line with the SLA, ISPs must consider the SLA constraints directly when placing VNFs and routing service requests.In this paper, we present a VNF placement and routing algorithm based on the column generation method which iterates between generating improving paths, and optimising the placement of the VNFs and routing of traffic given the generated paths.SLA constraints are modelled as soft constraints for which violation incurs a cost, the sum of which is minimised.Unlike prior approaches, we consider the throughput, latency and availability SLA constraints.We validate our approach on a number of mobile edge cloud (MEC) networks using 3 realistic network slice scenarios.We show that our approach can find near optimal solutions (±10% of optimal value) to realistic sized scenarios (up to 28 vertices, 41 edges and 700 service requests) within a reasonable time-frame (< 1 hour).