An Extensible Quantum Network Simulator Built on ns-3: Q2NS Design and Evaluation

TL;DR

Q2NS is a modular and extensible quantum network simulator, built on top of ns-3, designed to seamlessly integrate quantum-network primitives with ns-3's established classical protocol stack, and offers a flexible, open, and scalable simulation platform for advancing Quantum Internet research.

Abstract

As quantum networking hardware remains costly and not yet widely accessible, simulation tools are essential for the design and evaluation of quantum network architectures and protocols. However, designing a scalable and computationally efficient quantum network simulator is intrinsically challenging: i) quantum dynamics must be emulated on classical computing platforms while capturing the stateful and non-local nature of entanglement, a quantum resource without any classical networking analog; ii) quantum networking is inherently hybrid, as protocol execution also fundamentally depends on classical signaling. This makes a tight and faithful co-simulation of quantum operations and classical message exchanges a core requirement. In this light, we present Q2NS, a modular and extensible quantum network simulator, built on top of ns-3, designed to seamlessly integrate quantum-network primitives with ns-3's established classical protocol stack. Q2NS adopts a modular architecture that decouples protocol control logic from node- and channel-level operations, enabling rapid prototyping and adaptation across heterogeneous and evolving Quantum Internet scenarios. Q2NS natively supports multiple quantum state representations through a unified interface, allowing interchangeable state-vector, density-matrix, and stabilizer backends. We validate Q2NS through realistic use-case studies and comprehensive benchmarks, demonstrating superior computational efficiency over representative state-of-the-art alternatives, while preserving modeling flexibility. Finally, we provide a dedicated visualization tool that jointly captures physical and entanglement-enabled connectivity and supports entangled-state manipulations, facilitating an intuitive interpretation of entanglement dynamics and protocol behavior. Q2NS offers a flexible, open, and scalable simulation platform for advancing Quantum Internet research.

Figures (9)

• Figure 1: Graphical representation of the Q2NS simulation environment, illustrating the main entities (NetController, QNodes, QChannels) and their components. The diagram highlights the simulator's modular architecture, the separation of concerns and the interactions between modules and their hierarchies within the simulation framework.
• Figure 2: Visualization Flow: Q2NSViz$\rightarrow$ Trace File $\rightarrow$ Viewer. The generated logs correspond to an "at-the-source" entanglement generation scheme, where Alice creates an entangled pair and transmits one half to Bob.
• Figure 3: Simulation time (a) and memory (b) scaling of cluster state preparation and evaluation for qns-3 versus Q2NS (ket, density matrix, and stabilizer backends). Insets show the full range of the stabilizer backend and its best fit models, which extends beyond the axis limits that were chosen to clearly depict the other backends. Axes are logarithmic. Error bars represent $\pm 1$ standard deviation.
• Figure 4: Comparison of qns-3 and Q2NS for entanglement swapping chains as the number of network nodes increases, demonstrating Q2NS's higher performance. Axes are logarithmic. Error bars denote $\pm 1$ standard deviation. Note that qns-3 CFA-GR has a maximum size of $N = 128$ owing to the 1-hour time cutoff for single run data collection. For this reason the ratio plots on the bottom only extend to $N = 128$.
• Figure 5: Comparison of both versions of qns-3 CFA and Q2NS for an entanglement swapping chain with increasing number of network nodes, broken down by mean configuration and simulation time. The x axes are logarithmic. Note that in (b), qns-3 CFA-GR data is not shown as it is orders of magnitude larger, making it impossible to plot on a linear scale with the rest of the data. qns-3 CFA-GR also has a maximum size of $N = 128$ owing to the 1-hour time cutoff for single run data collection.