Dynamic resource allocation for URLLC and eMBB services in NFV-MEC 5G networks

Abstract

The Fifth Generation of mobile networks (5G) seeks to support a diversity of applications categorized into three types: enhanced Mobile Broadband (eMBB), massive Machine Type Communications (mMTC), and Ultra Reliable Low Latency Communications (URLLC), being their coexistence a major challenge. Multi-access Edge Computing (MEC), Network Function Virtualization (NFV) and Network Slicing (NS) emerge as complementary paradigms that shall support both eMBB and URLLC by offering fine-grained on-demand distributed resources closer to the User Equipment (UE) with a shared utilization of physical infrastructure. In this work, we have addressed the combination of MEC, NFV, NS and dynamic virtual resource allocation in order to overcome the problem of resource dimensioning in the network edge core. Thus, we have designed an analytical model to evaluate how requests are managed by the virtualization resources of a single MEC node, with a primary focus on meeting the requirements of both eMBB and URLLC services. We proposed a CTMC-based model to characterize dynamic virtual resource allocation and incorporated five performance metrics, which are relevant not only for URLLC and eMBB services (e.g., availability and response time) but also for service providers (e.g., power consumption), integrating practical factors like resource failures, service prioritization, and setup (repair) times into the formulation. This model enables an understanding of how the 5G network core behaves in serving different service categories by applying service prioritization to efficiently share processing resources. Some of our key findings include the idea that higher eMBB arrival rates decrease availability and increase response times up to 300 ms, while URLLC availability remains stable. Moreover, the container setup rates and failure rates substantially affect both availability and response times, with higher setup rates enhancing availability by up to 30% and reducing response times by 60%. Also, the number of containers emerges as a significant factor, enhancing both availability and response times, while buffer sizes mainly impact response times. In brief, our work advances in the current state of the art of the MEC-NFV domain by providing valuable insights for the design of MEC-NFV architecture, business models, and mechanisms to address communication constraints.