Towards End-to-End Application Slicing in Multi-access Edge Computing systems: Architecture Discussion and Proof-of-Concept

TL;DR

The paper addresses end-to-end latency guarantees in MEC-enabled 5G networks by proposing an end-to-end application-centric slicing framework. It introduces Application Slice (APS), Application Service (AS), and Application Component Functions (ACFs), and defines MAPSS as MEC application slice subnets, along with two deployment models. A two-layer MEC architecture (MEC Owner and MEC Customer) and a cross-domain management approach are proposed to enable multi-tenant, isolated MEC environments. A Kubernetes-based PoC demonstrates isolated MAPSS instances and analyzes performance isolation under resource constraints and shared resources, highlighting both feasibility and practical challenges. The work points to future directions in MEC relocation, SLA translation, and secure, efficient resource sharing across slices.

Abstract

Network slicing is one of the most critical 5G pillars. It allows for sharing a 5G infrastructure among different tenants leading to improved service customisation and increased operators' revenues. Concurrently, introducing the Multi-access Edge Computing (MEC) into 5G to support time-critical applications raises the need to integrate this distributed computing infrastructure to the 5G network slicing framework. Indeed, end-to-end latency guarantees require the end-to-end management of slice resources. For this purpose, after discussing the main gaps in the state-of-the-art with regards to such an objective, we propose a novel slicing architecture that enables the management and orchestration of slice segments that span over all the domains of an end-to-end application service, including the MEC. We also show how this general management architecture can be instantiated into a multi-tenant MEC infrastructure. A preliminary implementation of the proposed architecture focusing on the MEC domain is also provided, together with performance tests to validate the feasibility and efficacy of our design approach.

Figures (18)

• Figure 1: Exemplary illustration of the relationships between communication services, network slices, network slice subnets and resources/network functions
• Figure 2: High Level Roles in Network Slice Framework
• Figure 3: 3GPP network slice management architecture
• Figure 4: MEC reference architecture (based on MEC003).
• Figure 5: End-to-end network latency of a NSI that include MEC according to MEC024.