Tenant-First MEC Orchestration with Serverless Abstractions and Container Checkpoint/Restore
Coutinho, P.
;
Escaleira, P.
; Rocha, D.
; Almeida, J. A.
;
Gomes, D.Gomes
;
Aguiar, R.
Tenant-First MEC Orchestration with Serverless Abstractions and Container Checkpoint/Restore, Proc International Conference on Network and Service Management CNSM, Madrid, Spain, Vol. , pp. - , October, 2026.
Digital Object Identifier:
Abstract
The European Telecommunications Standards Institute (ETSI) Multi-access Edge Computing (MEC) architecture defines management functions for application lifecycle control, placement, scaling, event routing, and mobility. In practice, these functions are often implemented through layered MEC and Network Function Virtualization (NFV) Management and Orchestration (NFV-MANO) orchestration stacks, increasing operational complexity and lifecycle latency. This article proposes a tenant-first MEC orchestration architecture that exposes an ETSI-aligned declarative interface while realizing the underlying management functions through serverless primitives. Tenants specify application requirements such as placement constraints, event filters, scaling policies, and checkpoint behavior, while a MEC orchestrator translates them into serverless resources. To mitigate serverless cold start latency, the proposed architecture checkpoints idle containers and restores them on demand. We implement the system on Knative and evaluate it on a heterogeneous Cooperative Intelligent Transport Systems (C-ITS) testbed with Roadside Units (RSUs) and Virtual Machines (VMs). Results show that checkpoint-based restore reduces activation latency by up to 8.2× compared with cold starts, while checkpoint-assisted migration is up to 5.9× faster than cold-start redeployment. The end-to-end C-ITS pipeline achieves an average latency of 18.9 ms.