Surfing the SWAVES: Lifecycle-aware Service Placement in MEC

In Multi-access Edge Computing (MEC) networks, mobile users can offload virtual network functions (VNFs) to edge clouds (ECs) to meet stringent latency requirements through proximity. Under these requirements mobile users require VNFs to migrate across edge nodes as handovers occur. This process is non-trivial when ECs have limited resources and VNFs instantiation involves time-consuming operations, i.e., lifecycle operations. Ignoring lifecycle during placement decisions leads to severe service disruptions, as VNFs may not be ready to serve users when needed. To mitigate this issue, VNFs can be proactively instantiated at anticipated future locations, trading improved service continuity for increased resource consumption. Since exact future user locations are generally unknown, placement decisions must rely on connectivity forecasts, which are inherently imperfect and introduce an additional accuracy-resource tradeoff. We propose SWAVES, a lifecycle-aware VNF placement framework that explicitly accounts for instantiation delays and forecasting uncertainty. Simulation results show that by considering lifeycle operations and future forecasts, SWAVES closes the gap with an oracle approach that knows connectivityߣs ground truth. It also improves by orders of magnitude the number of unsuccessful packets with respect to other approaches ignoring lifecycle operations.