Design the network.
Compile the circuit. See what it costs.
QNEST is a graphical toolkit for optical quantum data centers. Build a topology out of QPUs,
switches and fiber, distribute a quantum circuit across it, schedule the remote gates, and run the
whole thing on established tools — MQT Bench, pytket-dqc, Qoala and NetSquid — wired into one
workflow, with the optical switching protocols and the interface on top.
The tools for quantum data center research don't talk to each other.
Circuit frameworks, distributed compilers and network execution environments each solve one layer
well. Nothing connects them, so the question that matters most — how the optical network changes
what the application actually achieves — falls between them.
Circuit frameworks
Qiskit · PyTket · QASM
Strong circuit design and compilation, but a single-device view. The interconnect isn't part of the model.
Distributed compilers
pytket-dqc
Partition circuits across modules and allocate resources. On their own terms they work well, but the communication they assume stays abstract — nothing ties the partition back to real fiber.
Network execution
Qoala · NetSquid
Rigorous node scheduling and protocol-level modelling of quantum communication, with no notion of the circuit that produced the traffic.
QNEST
The connective tissue
Rather than reimplementing any layer, QNEST wires these tools into one workflow and adds what was missing: optical network design, switching protocols, and a graphical environment over the whole run.
Built on
Established tools, not our own assumptions.
Circuit scheduling and entanglement generation used to rest on simple in-house heuristics. They
now rest on published, validated frameworks, so results carry the weight of the tools underneath
them.
Workloads
MQT Bench
Pick the benchmark circuit and the number of qubits for each simulation.
Distribution
pytket-dqc
Compiles and distributes the circuit over the quantum network you defined.
Scheduling
Qoala
Node-level execution and scheduling for quantum internet nodes, from TU Delft. arXiv:2502.17296
Execution
NetSquid
Discrete-event simulation of the quantum network the schedule runs on.
The workflow
Five stages, in order, end to end.
Each stage hands its output to the next one. Pick a stage to see what it does.
Pick the workload
Circuits come from MQT Bench: choose the benchmark and the number of qubits for the run. Gate
and process durations are set here too, so the timing model matches the hardware you have in
mind.
MQT Bench circuits, selected by benchmark and qubit count
QASM import and circuit inspection
Per-gate and per-process durations, in nanoseconds
Remote gate, crossing the optical link Local gate
Place QPUs, switches and fiber
The optical layer is a design object, not a hidden abstraction. Drag QPUs, beam splitters,
switches, quantum memories and links onto the canvas and wire them together.
Interactive topology editor, plus an editable table view
Right-click any component to set qubit count, T1 and T2, gate error, readout error, noise model
Pull calibration data from a real backend by name
Export and import a topology as JSON
QPU connectivity handed straight to the distributed compiler
Topology editorComponent parameters
Distribute across QPUs
pytket-dqc compiles the circuit against the network you just defined and distributes it over
the QPUs. What that costs becomes explicit: which gates turn remote, what they depend on, and
how much entanglement they will consume.
Distributed compilation with pytket-dqc
Partitioning driven by the QPU connectivity you defined
Remote-gate dependencies and entanglement demand, made explicit
Every cut edge becomes a remote gate, and every remote gate has to be paid for in entanglement.
Order the operations in time
The distributed circuit is passed to the QPUs and scheduled using the strategies defined in
Qoala and NetSquid, rather than in-house heuristics. The result is drawn as an interactive
Gantt chart and exported as standalone HTML.
Scheduling strategies from Qoala and NetSquid, both from TU Delft
Local operations, remote gates and communication events on one timeline
Hover any bar for timing, duration, resource and process detail
Interactive HTML output you can keep, share or drop into a talk
Timed circuit scheduleGate timings
Switch, execute and measure
Our switching protocols are applied on top, for either an all-photonic or a memory-assisted
architecture, and the run plays out over the configured optical infrastructure. Results come
back as detailed tables — every event, delay and resource in view.
All-photonic and memory-assisted architectures, side by side
Entanglement generation, distribution, storage and consumption
Communication delays and hardware imperfections
Tabulated results down to the individual event, with plots for final figures
A single remote gate, as the event engine sees it.
Capabilities
What's in the box.
Network-centric design
QPU placement, optical switching, quantum memories and links are all configurable objects, so architectural questions can be asked directly.
Realistic device models
Qubit counts, coherence times, gate and readout errors, noise models, and calibration data fetched from named backends.
Distribution via pytket-dqc
Circuits compiled and distributed across QPUs by an established tool, producing explicit remote-gate and entanglement requirements.
Scheduling you can cite
Gantt views tie local operations, remote gates and communication to one clock, scheduled by Qoala and NetSquid rather than in-house heuristics.
Two switching architectures
The switching protocols run over either an all-photonic or a memory-assisted design, so the two can be compared on identical workloads.
Cross-layer analysis
Application, compilation, scheduling and network stack in one place — physical-layer choices trace through to application results.
Graphical, not scripted
No command line required. The barrier to entry for distributed quantum computing research drops to installing an app.
Standalone and cross-platform
Runs on ordinary Windows, macOS and Linux workstations, which keeps experiments easy to reproduce and share.
Readout
What a run tells you.
Five numbers that change when the network changes. Results are reported as detailed tables, so
every event, delay and allocation stays visible — useful while a design is still being debugged,
and easy to turn into plots once the numbers settle.
Entanglement fidelity
Quality of the pairs actually delivered to remote gates, after distribution and storage.
Job execution time
Wall-clock length of the distributed job, JET, including everything the network adds.
Communication overhead
Time and resources spent moving quantum information between QPUs rather than computing.
Entanglement-resource use
How hard the entanglement generation and memory resources are being worked.
Penalty delay
Waiting time caused by a remote gate arriving before an entangled pair is available.
Architecture
How the stages fit together.
From circuit specification and optical-network configuration through distributed compilation,
scheduling, execution and performance evaluation — one path, one environment.
QNEST workflow and software architecture.
See it live
QNEST is a demonstration at ECOC 2026.
Come and drive it yourself. Change the topology, the noise or the scheduling policy, and watch the
application-level numbers move.
Department of Electrical Engineering, Chalmers University of Technology, Gothenburg, Sweden ·
Smart Internet Lab, University of Bristol, United Kingdom elyasi@chalmers.se
Cite
If QNEST helps your work, cite the paper.
@inproceedings{elyasi2026qnest,
title = {QNEST: A GUI-Driven Interactive Framework for End-to-End
Simulation of Optical Quantum Data Centers},
author = {Elyasi, Seyed Navid and Bahrani, Sima and Wang, Rui and
Simeonidou, Dimitra and Monti, Paolo and Lin, Rui},
booktitle = {European Conference on Optical Communication (ECOC)},
year = {2026}
}
The demo paper is available as a PDF. The abstract was accepted
for demonstration D22 at ECOC 2026 under the project's former name, GsOQDC.
Supported by
Supported by the Swedish Research Council (VR) and the UK EPSRC Integrated Quantum Networks Hub
(EP/Z533208/1).
