Uma abordagem para o controle de elasticidade dinâmico voltada a sistemas cloud-network definidos por slices

Abstract

In the last decade, telecommunications operators have adopted network slicing technology in their systems, applying a layer of virtualization over the physical infrastructure, allowing them to be sliced into different portions to be made available as a service to their customers. Softwarisation and cloudification are key technologies in the architecture of 5G systems, and key enablers for efficiently supplying network slices. However, the integration of these technologies in a coordinated way on the same virtualized infrastructure still presents itself as a challenging task. In this scenario, recent approaches aim to extend the concept of network slicing to cloud, with the perspective of allowing slicing in cloud infrastructure and re-orchestrating in a coordinated way, thus characterizing slice-defined cloud-network systems. In this new scenario, an efficient approach to the control of resource elasticity is a fundamental factor to ensure rational and efficient use of cloud and network resources, thus guaranteeing the quality of the services offered by these systems, as well as allowing a significant reduction of capital (CAPEX) and operational (OPEX) expendidures of these systems. While the state of the art presents a great majority of works that propose a stochastic approach to the control and allocation of resources (eg, bandwidth, CPU, memory, mass storage, and others), we introduce a mechanism to elasticity control in slice-defined cloud-network systems that presents greater efficiency in the allocation of resources in critical moments of the system, main disadvantage of stochastic models. Thus, the proposed SLOTS (elaSticity in cLOud-neTwork Slices) approach has as main characteristics an end-to-end vision in multi-domain environments and the knowledge of critical situations driven by resource depletion. A SLOTS prototype was implemented and its preliminary results suggest that its approach is a viable solution to provide resource elasticity in environments defined by cloud-network slices under critical resource conditions and high demand.