A Lightweight, Open-Source Network Digital Twin Blueprint for IoT-Edge Environments

arXiv:2606.24853 · cs.NI · Submitted 2026-06-23 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Computer Science. The day in computer science: software, systems and theory on arXiv.

Tom: I'm Tom.

Jane: And I'm Jane, here with Lu, Meng and our in-house large language model, Lalam.

Tom: We're looking at the paper: "A Lightweight, Open-Source Network Digital Twin Blueprint for IoT-Edge Environments".

Jane: Here's the gist of the paper. Validating network configurations and testing failure scenarios in IoT-edge-cloud environments without disrupting live infrastructure remains an open operational challenge.

Tom: This paper presents a low-cost, fully open-source Network Digital Twin (NDT) for IIoT edge deployments, built on Containerlab, Open vSwitch, ONOS, and a Prometheus/Grafana observability stack.

Jane: The framework integrates container-native topology emulation, SDN-driven traffic engineering, and real-time telemetry in a single deployable artefact.

Tom: NDT Architecture and Components The NDT is structured around three functional planes (Fig.

Paper summary: Tom: So, what’s the main takeaway here? The paper argues that this NDT framework can effectively validate network configurations and test failure scenarios in IoT-edge-cloud environments safely. It claims they have built a low-cost, fully open-source tool to do this without disrupting live infrastructure.

Jane: What does that actually mean for the people who are working in IIoT deployments? It seems like it gives them a way to stress test their network setups in a controlled way before they push anything live.

Lu: The architecture is structured into three planes, right? There's an emulation plane that builds the virtual topology and injects channel impairments, then there's the control plane managed by ONOS for SDN traffic engineering, and finally the observability plane that collects data from both the NDT and the physical network.

Meng: I see what you’re saying about those planes. So they are using Containerlab to create Linux containers that mirror a testbed with six Raspberry Pi nodes and an access point, right? That's a lot of specific hardware modeling in software.

Lalam: Exactly. In the emulation plane, they define the network topology in a file called ndt-topology yaml, which is a Containerlab manifest. Then they apply traffic impairments at the access point facing virtual interfaces using tc netem for delay and jitter, plus a Token Bucket Filter shaper to simulate real channel issues.

Tom: And on the control plane side, ONOS connects to each Open vSwitch bridge using OpenFlow version one point three over TCP port six six five three. That’s how they program the traffic flow dynamically based on what ONOS tells it to do.

Jane: But how do they prove it works? The paper says they validate this against a far edge small testbed deployed in the fortiss IIoT Labs, and they found strong distributional convergence on RTT median with a delta of zero point four milliseconds and UDP throughput with a delta of zero point zero three megabits per second.

Lu: Those numbers suggest that even when simulating impairments, the network behavior matches reality pretty closely for those specific metrics. The paper also notes that remaining differences in TCP throughput and packet loss are due to virtualization artifacts, which they’ve identified and provided ways to fix them.

Meng: So they found these small discrepancies in TCP throughput and packet loss, but they figured out what's causing them—the virtualization stuff—and showed how to handle it. That’s important because real-world deployments always have those kinds of weird edge cases.

Lalam: It means this isn't just a theoretical model; it’s actually running against some real physical hardware, and they aren't pretending everything is perfect when it isn't. That adds a lot of credibility to the whole framework.

Conclusion: Tom: So looking at the whole "A Lightweight, Open-Source Network Digital Twin Blueprint for IoT-Edge Environments" paper, the authors are really focusing on providing a low-cost way to validate networks in these complex IIoT edge environments safely and openly.

Jane: I think what matters here is that they’ve made this tool fully open source. That means anyone can use it to test their own setups without being locked into expensive, proprietary platforms. It lowers the barrier for doing serious testing on the edge.

Lu: The implication is that we can move away from just simulating simple delay models and actually test how real SDN control planes interact with containerized network stacks in a way that matches what happens in the physical world.

Meng: For someone trying to deploy models, this could mean they have a better sandbox for testing their communication layer interactions before deploying to the actual edge devices. It’s about finding those bugs early on in a controlled environment.

Lalam: The title itself points to how light and open-source it is, which is key for adoption in industrial settings where you need transparency. It’s a blueprint, not just a single experiment; it shows how to build this whole system.

Tom: So the main point isn't just that they can test networks; it’s that they created a complete, integrated digital twin artifact that connects emulation, SDN control, and observation in one piece.

Jane: It gives engineers a concrete blueprint for how to approach testing these challenging IIoT environments using accessible tools like Containerlab and ONOS.

Lu: The paper suggests that the next step is figuring out those open engineering challenges they mentioned, particularly how to faithfully reproduce the physical layer behavior across all those different containerized nodes.

Meng: I wonder what it would cost to run this at scale in a real production setting? That’s where I’d want to know if it actually holds up beyond this small testbed validation.

Lalam: The paper does mention that the integration between Kubernetes and Open vSwitch creates complexity, which is a good piece of honesty about the limits of the current design.

Tom: So in short, this blueprint gives people a practical starting point for creating their own digital twin for edge testing without needing massive budgets or proprietary software.

fortiss GmbH · ISTEC

cs.NI

Submitted: 2026-06-23

Updated: 2026-11-10

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 80/100

The gist: Validating network configurations and testing failure scenarios in IoT-edge-cloud environments without disrupting live infrastructure remains an open operational challenge.

Key concepts

Network Digital Twin (NDT)
A virtual model that perfectly mimics a physical network. This twin is built using containers and software like Open vSwitch to replicate the exact layout, hardware, and traffic behavior of a real IoT edge deployment in a safe, virtual environment.
Containerlab
A tool used to quickly set up complex containerized networks. It allows researchers to easily provision the necessary Linux containers that represent the physical nodes and access points of an IoT testbed for simulation purposes.
SDN-driven Traffic Engineering
Using Software-Defined Networking (SDN) managed by ONOS to programmatically control network traffic flow. This means the system can intelligently reroute data paths in real-time to improve performance or handle failures automatically, leading to faster recovery times.
Prometheus/Grafana Observability
A system used to monitor and visualize network performance data. It collects metrics from both the simulated NDT and the actual physical network, allowing for direct comparison of results using standard tools like PromQL.

Terminology

Summary

Validating network configurations and testing failure scenarios in IoT-edge-cloud environments without disrupting live infrastructure remains an open operational challenge. p. 0

Where the paper says it

Sofia fortiss GmbH Munich, Germany sofia@fortiss. org Abstract—Validating network configurations and testing failure scenarios in IoT-edge-cloud environments without disrupting live infrastructure remains an open operational challenge.

This paper presents a low-cost, fully open-source Network Digital Twin (NDT) for IIoT edge deployments, built on Containerlab, Open vSwitch, ONOS, and a Prometheus/Grafana observability stack. p. 0
Where the paper says it

org Abstract—Validating network configurations and testing failure scenarios in IoT-edge-cloud environments without disrupting live infrastructure remains an open operational challenge. This paper presents a low-cost, fully open-source Network Digital Twin (NDT) for IIoT edge deployments, built on Containerlab, Open vSwitch, ONOS, and a Prometheus/Grafana observability stack.

The framework integrates container-native topology emulation, SDN-driven traffic engineering, and real-time telemetry in a single deployable artefact. p. 0
Where the paper says it

This paper presents a low-cost, fully open-source Network Digital Twin (NDT) for IIoT edge deployments, built on Containerlab, Open vSwitch, ONOS, and a Prometheus/Grafana observability stack. The framework integrates container-native topology emulation, SDN-driven traffic engineering, and real-time telemetry in a single deployable artefact.

NDT Architecture and Components

The NDT is structured around three functional planes (Fig. 1). The emulation plane instantiates the virtual topology and injects channel impairments. p. 2

Where the paper says it

Mapping to the IRTF Reference Model The NDT is structured around three functional planes (Fig. 1): the emulation plane instantiates the virtual topology and injects channel impairments; the control plane provides programmable SDN management via ONOS; and the observability plane collects, stores, and visualizes telemetry from both the NDT and the physical CODECO network.

The control plane provides programmable SDN management via ONOS. p. 2
Where the paper says it

Mapping to the IRTF Reference Model The NDT is structured around three functional planes (Fig. 1): the emulation plane instantiates the virtual topology and injects channel impairments; the control plane provides programmable SDN management via ONOS; and the observability plane collects, stores, and visualizes telemetry from both the NDT and the physical CODECO network.

The observability plane collects, stores, and visualizes telemetry from both the NDT and the physical CODECO network. p. 2
Where the paper says it

Mapping to the IRTF Reference Model The NDT is structured around three functional planes (Fig. 1): the emulation plane instantiates the virtual topology and injects channel impairments; the control plane provides programmable SDN management via ONOS; and the observability plane collects, stores, and visualizes telemetry from both the NDT and the physical CODECO network.

Emulation Plane Details

The network topology is declared in ndt-topology.yaml, a Containerlab manifest that provisions Linux containers mirroring the six CODECO testbed Raspberry Pi nodes, and the access point (AP). p. 2

Where the paper says it

Emulation Plane The network topology is declared in ndt-topology. yaml, a Containerlab manifest that provisions Linux containers mirroring the six CODECO testbed Raspberry Pi nodes, and the access point (AP).

Traffic impairment is applied at AP-facing virtual interfaces using tc netem (delay, jitter, loss) in combination with a Token Bucket Filter (TBF) shaper. p. 2
Where the paper says it

3, allowing flow rules to be installed, modified, and deleted programmatically. Traffic impairment is applied at AP-facing virtual interfaces using tc netem (delay, jitter, loss) in combination with a Token Bucket Filter (TBF) shaper (bandwidth cap), encapsulated in the wifi-profile.

The four pre-defined profiles are listed in Table II. p. 2
Where the paper says it

Traffic impairment is applied at AP-facing virtual interfaces using tc netem (delay, jitter, loss) in combination with a Token Bucket Filter (TBF) shaper (bandwidth cap), encapsulated in the wifi-profile. The four pre-defined profiles are listed in Table II.

Control and Observability Mechanisms

ONOS connects to each OVS bridge via OpenFlow 1.3 on TCP port 6653. p. 3

Where the paper says it

7) connects to each OVS bridge via OpenFlow 1. 3 on TCP port 6653.

Prometheus scrapes two jobs simultaneously at 15-second intervals: ndt-nodes and real-nodes. p. 3
Where the paper says it

Observability Plane Node Exporter agents are embedded in every emulated container via Containerlab exec directives, exposing perinterface byte/packet counters, CPU usage, memory, and system load on port 9100. Prometheus scrapes two jobs simultaneously at 15-second intervals: • ndt-nodes: the six emulated NDT containers • real-nodes: the six physical CODECO Raspberry Pi nodes This unified scrape configuration enables direct side-by-side comparison using standard PromQL.

This unified scrape configuration enables direct side-by-side comparison using standard PromQL. p. 3
Where the paper says it

Observability Plane Node Exporter agents are embedded in every emulated container via Containerlab exec directives, exposing perinterface byte/packet counters, CPU usage, memory, and system load on port 9100. Prometheus scrapes two jobs simultaneously at 15-second intervals: • ndt-nodes: the six emulated NDT containers • real-nodes: the six physical CODECO Raspberry Pi nodes This unified scrape configuration enables direct side-by-side comparison using standard PromQL.

Experimental Validation and Findings

The framework was validated against a far Edge small testbed deployed in the fortiss IIoT Labs. Key findings include:

  1. Four calibrated Wi-Fi impairment profiles show that it is feasible to reproduce in a sound way the Wi-Fi channel degradation.

  2. ONOS-driven SDN traffic engineering achieves subsecond failure recovery spanning three orders of magnitude of latency improvement.

  3. A 10-run short-term controlled session demonstrates pointconvergence on RTT average, UDP jitter, and UDP packet loss.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improved AI systems can perform network configuration validation against physical environments by using a low-cost, fully open-source Network Digital Twin (NDT) built on Containerlab, Open vSwitch, ONOS, and Prometheus/Grafana to provide a safe sandbox for experimentation.

  2. The NDT framework allows for the creation of calibrated Wi-Fi impairment profiles derived from empirical CODECO measurements via a reusable wifi-profile.sh toolchain, enabling the AI system to accurately model and test network behavior under specific congestion scenarios.

  3. The system can execute a five-experiment evaluation campaign, including baseline characterization, congestion impact, SDN traffic engineering, resilience testing, and NDT-vs-physical fidelity validation to systematically assess network performance in complex IoT-edge deployments.

  4. The AI system can identify and mitigate specific virtualization artifacts such as the TBF segment starvation artefact observed in Containerlab/OVS environments when iPerf3 uses default MTU-sized transfers, improving the fidelity of throughput measurements.

  5. The system can dynamically adjust network paths by leveraging ONOS's Intent Framework, enabling SDN traffic engineering that achieves subsecond failure recovery spanning three orders of magnitude of latency improvement.

Abstract

IoT-Edge-Cloud infrastructures are increasingly heterogeneous, and experimenting across the continuum is difficult and risks disrupting live systems. Network Digital Twins (NDT) are a pragmatic answer, yet practitioners lack a structured account of three things: how to build an NDT from open tools, how to model wireless impairments, and how to understand whether the twin is faithful to the physical network, or not. This tutorial introduces NDT through the IRTF reference model and walks through a lightweight open-source blueprint built on Containerlab, Open vSwitch, Open Network Operating System (ONOS), and Prometheus/Grafana. It then explains how to define wireless impairment profiles (static, model-based, trace-driven, AI-generated, and live), and what each approach can and cannot reproduce. Finally, it presents a workflow to validate the twin against the physical network. A proof-of-concept Edge case study illustrates the procedure, including the pitfalls encountered and the lessons learned. Directions for future research are outlined.

Related papers