Serverless rApps at the Network Edge: An Experimental Evaluation of Event-Driven Non-RT RIC Functions on Kubernetes-Based MEC
The O-RAN Non-RT RIC is hosting an increasing number of rApps for data-driven network management, but current deployments run them all as always-on microservices, even when many are only needed intermittently. This leads to inefficient use of limited MEC resources. This paper examines whether serverless execution is a practical alternative for edge deployment of rApps. We design an event-driven serverless rApp framework and evaluate it on a MEC testbed with three workloads: periodic energy saving, event-triggered fault remediation, and semi-continuous traffic steering. We compare always-on, fully serverless, and hybrid deployments in terms of cold-start latency, control-loop response, burst autoscaling, and 24-hour resource usage. Results show that serverless deployment cuts resource consumption by 35–50% for sporadic workloads while keeping latency within Non-RT RIC timescales. However, high-frequency rApps suffer excessive delay under scale-to-zero policies. Based on these results, we advocate a hybrid deployment model and provide guidelines for matching rApp workload profiles to execution strategies.
Authors
- Roberto Riggio (ORCID: https://orcid.org/0000-0002-8329-2779)
Institutions
- Marche Polytechnic University (IT)
Publication Details
- Journal
- Journal of Network and Systems Management
- Published
- 2026-11-21
- DOI
- https://doi.org/10.1007/s10922-026-10111-z
- Primary Topic
- Software-Defined Networks and 5G
- Type
- article
- Field-Weighted Citation Impact
- 0.00