Resource allocation for URLLC in NFV-MEC

• Autor(a) principal: FALCÃO, Marcos Rocha de Moraes
• Data de Publicação: 2022
• Tipo de documento: Tese
• Idioma: eng
• Título da fonte: Repositório Institucional da UFPE
• Texto Completo: https://repositorio.ufpe.br/handle/123456789/45633
• topic: Rede de computadores e sistemas distribuídos; Computação de borda multiacesso; Virtualização de funções de rede; Alocação de recursos

Multi-access Edge Computing (MEC) and Network Function Virtualization (NFV) emerge as complementary paradigms that shall support Ultra-reliable Low Latency Communication (URLLC) by offering fine-grained on-demand distributed resources closer to the User Equip- ment (UE), thus mitigating physical layer issues. On the other hand, the adoption of the NFV-MEC inevitably raises deployment and operation costs. We have addressed the combina- tion of MEC, NFV and dynamic virtual resource allocation in order to overcome the problem of resource dimensioning in a special scenario were MEC infrastructure is mounted over Un- manned Aerial Vehicles (UAVs) in the context of URLLC. First, a Continuous-time Markov Chain (CTMC)-based model was proposed to characterize dynamic virtual resource allocation in the MEC node together with four performance metrics that are both relevant for URLLC applications (e.g., reliability and response time) and for service providers (e.g., availability and power consumption). In order to yield the model more practical, the effect of virtual host resource failures, setup (repair) times and processing overheads were embedded into the for- mulation, since they may significantly affect the stringent requirements of URLLC applications. Moreover, a multi-objective problem related to MEC-enabled UAV node dimensioning in terms of virtual resources (VMs, containers and buffer positions) was formulated. In this context, the compromise between on-board computation resources and the URLLC requirements become a great challenge since UAVs are limited due to their size, weight and power, which imposes a burden on the conventional Network Functions (NFs). Finally, an approach based on Genetic Algorithms (GA) was formulated to solve the dimensioning problem, with the proposed scheme achieving a better tradeoff in terms of availability, reliability, power consumption and response time compared to the commonly adopted approaches based on the First-fit strategy.