Dynamic VNF Placement in 5G Edge Nodes

Abstract

The ongoing transition into 5G networks is enabled in part by the combination of NFV (Network Function Virtualization) and MEC (Multi-access Edge Computing), two promising paradigms that allow executing ultra-low-latency network services on edge nodes, physically closer to the clients. However, orchestrating this complex distributed environment and especially provisioning services in a timely manner, in order to address the dynamic workload, remained a big challenge. In this paper we address this challenge and study ways to dynamically place network functions at edge nodes, across the network, in a way that maximizes client satisfaction, we measure this satisfaction by the number of clients that received their desired services in a manner that holds these required services low-latency demands. In order to balance between the dynamic workload and the non-negligible cost of replacing the functions at the edge, we partition the time into epochs and reassign VNFs (Virtual Network Functions) only at the beginning of each epoch. Our theoretical analysis, based on studying a simple variant of the online problem, shows that the data from the last epoch can provide guarantees on the expected performance. We then evaluate the actual performance of our algorithm based on extensive simulations over real data. The results indicate that our new algorithm can be deployed in a realistic 5G setting, generating an overall dynamic solution that outperforms currently used methods.