[Paper] Mini-SFC: A Comprehensive Simulation Framework for Orchestration and Management of Service Function Chains

Source: arXiv - 2512.11527v1

Overview

The paper presents Mini‑SFC, an open‑source, modular simulation framework designed to model, orchestrate, and manage Service Function Chains (SFCs) in cloud‑native networks. By combining lightweight numerical simulations with full‑stack container‑based emulation—and allowing the network topology to evolve on‑the‑fly—Mini‑SFC aims to bridge the gap between fast algorithm prototyping and realistic deployment testing.

Key Contributions

• Dual‑mode simulation – supports both fast, abstract numerical models and heavyweight Docker‑container based virtual environments within the same framework.
• Dynamic topology handling – lets users add, remove, or re‑configure network nodes and links while a simulation is running, mirroring real‑world scaling and failure scenarios.
• Unified solver interface – provides a standardized API for plugging in custom SFC placement, routing, or scaling algorithms, reducing boilerplate code.
• Modular architecture – separates topology, traffic generation, function virtualization, and performance measurement into interchangeable modules.
• Open‑source and developer‑friendly – comprehensive documentation, example scripts, and a plug‑and‑play Docker image lower the entry barrier for both researchers and industry engineers.

Methodology

Mini‑SFC is built around three core components:

1. Topology Engine – a graph‑based representation of the network (switches, servers, VNFs). Users can modify this graph at runtime via a RESTful control plane.
2. Simulation Core – runs either a numerical mode (using analytical models for latency, CPU, and bandwidth) or a container mode (spawning Docker containers that host real VNF images). The core synchronizes time steps, collects metrics, and forwards events to the solver.
3. Solver Layer – a pluggable Python/Java interface where developers drop in their SFC orchestration logic (e.g., placement heuristics, load‑balancing policies). The solver receives the current network state, makes decisions, and pushes configuration updates back to the topology engine.

The authors validate the framework by implementing a classic “first‑fit decreasing” placement algorithm and a reinforcement‑learning based optimizer, demonstrating that the same code can be evaluated in both simulation modes without modification.

Results & Findings

• Performance parity – In container mode, Mini‑SFC reproduces end‑to‑end latency and CPU utilization within 5 % of a real testbed, while numerical mode runs ≈30× faster for large‑scale topologies (10 k nodes).
• Rapid prototyping – Switching from a numerical to a container simulation for a given algorithm required only a single configuration change, cutting the validation cycle from days to hours.
• Dynamic reconfiguration – Experiments that added 20 % more VNFs during runtime showed the framework could re‑optimize placements in under 2 seconds, illustrating suitability for “elastic” SFC scenarios.

Practical Implications

• DevOps for network functions – Engineers can prototype placement or scaling policies locally with the fast numerical mode, then validate them on a realistic Docker‑based sandbox before rolling out to production.
• CI/CD pipelines – Mini‑SFC’s container mode can be integrated into automated testing suites, ensuring that new VNF images or orchestration scripts do not degrade performance.
• Edge and 5G deployments – The dynamic topology feature mirrors the churn of edge nodes (e.g., mobile base stations turning on/off), enabling developers to test resiliency and latency guarantees under realistic conditions.
• Education & training – Because the framework abstracts low‑level networking details while still offering a full‑stack view, it can be used in university labs or corporate training to teach SFC concepts without expensive hardware.

Limitations & Future Work

• Scalability ceiling – While the numerical mode scales to tens of thousands of nodes, the container mode is limited by host resources; large‑scale emulations still require a cluster of machines.
• VNF diversity – The current release ships with a handful of generic VNFs (firewall, NAT, DPI). Extending the library to include proprietary or hardware‑accelerated functions will need community contributions.
• Solver ecosystem – Only a few example solvers are provided; richer integration with existing SDN/NFV controllers (e.g., OpenDaylight, ONAP) is left for future development.
• Real‑time control plane – The REST API introduces latency that may not reflect ultra‑low‑latency control loops required in some 5G use cases; tighter integration with gRPC or message‑bus systems is planned.

Mini‑SFC positions itself as a practical bridge between theory and production for anyone building, testing, or optimizing Service Function Chains—making it a valuable addition to the toolbox of cloud‑native network engineers and researchers alike.

Authors

• Xi Wang
• Shuo Shi
• Chenyu Wu

Paper Information

• arXiv ID: 2512.11527v1
• Categories: cs.SE
• Published: December 12, 2025
• PDF: Download PDF