Why ncsim? · Documentation · Codespaces · Paper · Software DOI
Codespaces: The web UI should open automatically. If it does not, type
start-vizin the terminal, then open port 5173 from the Ports tab. Port 8000 is the internal API and is not the UI.
ncsim is a lightweight simulator for DAG scheduling over heterogeneous networked systems with multi-hop routing and realistic Wi-Fi interference modeling.
ncsim models compute nodes, network links with WiFi interference, and DAG task graphs. It produces detailed JSONL traces and JSON metrics for analysis.
Current manuscript sources and curated results are on the paper branch in
artifacts/arxiv-2605.01094/.
The earlier manuscript tree is archived separately as
arxiv-old/.
See wireless modes and fixed capture for
the optional model's scope and regression commands.
- Deterministic simulation: Same inputs + same seed = identical results
- 22+ SAGA static batch schedulers: HEFT, CPOP, Min-Min, Sufferage, and more; PEFT is added automatically with SAGA 2.1.0, alongside built-in round-robin and manual assignment
- Multi-hop routing: Direct, widest-path (max-min bandwidth), and shortest-path (min-latency)
- 802.11 WiFi PHY/MAC: Log-distance path loss, SNR-based MCS rate adaptation (802.11n/ac/ax)
- Interference models: Proximity, CSMA/CA clique-based, and CSMA/CA Bianchi (capture-aware)
- Fair bandwidth sharing when multiple transfers share a link
- Experiment scripts for interference verification and routing comparison
- Documentation: installation guide, quick start, architecture overview, and Wi-Fi interference model
Open ncsim in GitHub Codespaces for a ready-to-use environment with both the web UI and CLI. The UI starts automatically on port 5173, while the ncsim CLI is ready in the terminal. A demo simulation is also run during setup; inspect its raw scenario.yaml, trace.jsonl, and metrics.json files under results/codespaces-demo/.
Rerun the demo and analyze its trace from the terminal:
ncsim --scenario scenarios/demo_simple.yaml --output results/codespaces-demo
python analyze_trace.py results/codespaces-demo/trace.jsonl --gantt --timeline --tasksIf the UI does not open automatically, start or restart it with:
start-vizThen select the Ports tab at the bottom of Codespaces, hover over port 5173, and select the globe (Open in Browser).
Recommended: Clone the repository to get started. The repo includes example scenarios, experiment scripts, documentation, and the web visualization UI — all useful for learning and exploring ncsim:
git clone https://github.com/ANRGUSC/ncsim.git
cd ncsim
pip install -e .
# For development (includes pytest)
pip install -e ".[dev]"Alternatively, pip install anrg-ncsim installs just the core simulator and ncsim CLI. This is suitable if you want to use ncsim as a library in your own project and will write your own scenario YAML files. It does not include the example scenarios, experiment scripts, visualization UI, or documentation.
Requires Python 3.12+ and anrg-saga >= 2.0.4. The PyPI release of SAGA provides 22 directly compatible schedulers. To add PEFT as the 23rd scheduler, install SAGA 2.1.0 from its tagged source:
python -m pip install "anrg-saga @ git+https://github.com/ANRGUSC/saga.git@v2.1.0"ncsim --scenario scenarios/demo_simple.yaml --output results/Output:
results/trace.jsonl— event traceresults/metrics.json— summary metricsresults/scenario.yaml— copy of the input scenario
ncsim --scenario PATH --output DIR [options]
Options:
--seed N Random seed (default: from scenario or 42)
--scheduler ALGO SAGA scheduler, round_robin, or manual
--scheduler-option K=V
Scheduler constructor option (repeatable)
--routing ROUTING direct | widest_path | shortest_path
--interference MODEL none | proximity | csma_clique | csma_bianchi
--verbose Enable verbose logging
WiFi / RF options (for csma_clique or csma_bianchi):
--tx-power DBM Transmit power in dBm (default: 20)
--freq GHZ Carrier frequency in GHz (default: 5.0)
--path-loss-exponent N
Path loss exponent (default: 3.0)
--wifi-standard STD n | ac | ax (default: ax)
--rts-cts Enable RTS/CTS
scenario:
name: "Simple Demo"
network:
nodes:
- {id: n0, compute_capacity: 100, position: {x: 0, y: 0}}
- {id: n1, compute_capacity: 50, position: {x: 10, y: 0}}
links:
- {id: l01, from: n0, to: n1, bandwidth: 100, latency: 0.001}
dags:
- id: dag_1
inject_at: 0.0
tasks:
- {id: T0, compute_cost: 100}
- {id: T1, compute_cost: 200}
edges:
- {from: T0, to: T1, data_size: 50}
config:
scheduler: wba
scheduler_options:
alpha: 0.75
seed: 42Tasks can include pinned_to: node_id for use with --scheduler manual.
Run ncsim --help for the scheduler list provided by the installed SAGA version. SAGA scheduler options
currently available are fcp.priority_queue_size, gdl.dynamic_level,
smt.epsilon, smt.solver_name, and wba.alpha; all have SAGA defaults.
See scenarios/ for more examples including WiFi interference, multi-hop routing, and parallel spread topologies.
The paper-specific scenarios, recorded results, and reproduction instructions
for the IEEE MILCOM 2026 study are in
experiments/milcom26/.
Two standalone scripts for running structured experiments:
# Validate WiFi interference model against analytical predictions
python run_interference_verification.py
# Compare widest_path vs shortest_path routing on grid topologies
python run_routing_comparison.py
python visualize_routing_comparison.py # Generate plots from resultspython analyze_trace.py results/trace.jsonl --gantt --timeline --taskspython -m pytest tests/ -vAn extensive unit and integration suite covers the event queue, execution engine, scheduling, routing, Wi-Fi physics, visualization API, and acceptance criteria.
For a detailed overview, see the architecture documentation.
ncsim/ # Python package
├── main.py # CLI entry point
├── core/
│ ├── simulation.py # Main simulation loop
│ ├── event_queue.py # Priority queue with deterministic ordering
│ └── execution_engine.py
├── models/
│ ├── network.py # Node, Link, Network
│ ├── dag.py # DAG, Edge, Task
│ ├── routing.py # Direct, WidestPath, ShortestPath
│ ├── interference.py # Proximity, CSMA Clique, CSMA Bianchi
│ └── wifi.py # 802.11 PHY/MAC
├── scheduler/
│ ├── base.py # Scheduler interface
│ └── saga_adapter.py # SAGA static batch scheduler registry and adapter
└── io/
├── scenario_loader.py
├── trace_writer.py
└── results_writer.py
scenarios/ # Example scenario YAML files
tests/ # Unit and integration test suite
docs/ # MkDocs documentation source
ncsim includes an optional web UI (viz/) for interactive experiment configuration and result visualization. The viz is not included in the PyPI package — clone the repository to use it.
# Terminal 1: Backend API server
cd viz/server && pip install -r requirements.txt && python run.py
# Terminal 2: Frontend dev server
cd viz && npm install && npm run devOpen http://localhost:5173 to configure experiments, run simulations, and visualize results interactively. See viz/README.md for full documentation.
Build a scenario interactively — choose a scheduler, routing strategy, interference model, topology preset (line, star, ring, mesh, grid), and DAG preset (chain, fork-join, diamond, parallel). Edit nodes, links, and tasks in editable tables, then run the experiment with one click.
After running or loading an experiment, explore results across six tabs:
| Tab | Description |
|---|---|
| Overview | Makespan, task/transfer counts, node and link utilization bars |
| Network | Interactive D3 topology with node capacity and bandwidth labels |
| DAG | Task dependency graph with tasks colored by assigned node |
| Schedule | Gantt chart showing task execution windows across all nodes |
| Simulation | Animated replay: synchronized network view + live Gantt + event log |
| Parameters | Full scenario config inspector |
Overview — summary dashboard with node utilization
DAG — task dependency graph, colored by node assignment
Schedule — Gantt chart of task execution across nodes
Simulation — animated replay with live transfers, Gantt timeline, and event log
The simulation replay supports keyboard shortcuts: Space (play/pause), arrow keys (step events), +/- (speed 0.25x-10x), and keys 1-6 to switch tabs.
viz/ # Web visualization (React + FastAPI)
├── src/ # React frontend
├── server/ # FastAPI backend
└── public/ # Sample experiment runs
If you use ncsim in your research, please cite the paper and the software release:
@article{krishnamachari2026ncsimpaper,
author = {Krishnamachari, Bhaskar and Gutierrez, Maya and Coleman, Jared},
title = {ncsim: A Lightweight Simulator for Networked Edge Computing with Wireless Interference Modeling},
year = {2026},
url = {https://arxiv.org/abs/2605.01094},
note = {arXiv:2605.01094}
}
@software{krishnamachari2026ncsimsoftware,
author = {Krishnamachari, Bhaskar and Gutierrez, Maya},
title = {ncsim: A Lightweight Simulator for Networked Edge Computing with Wireless Interference Modeling},
version = {1.1.0},
year = {2026},
url = {https://github.com/ANRGUSC/ncsim},
doi = {10.5281/zenodo.19138224}
}This work was supported in part by Army Research Laboratory under Cooperative Agreement W911NF-17-2-0196.
See CONTRIBUTING.md for development setup, validation, and pull request guidance. Release history is recorded in CHANGELOG.md.
Bhaskar Krishnamachari, Maya Gutierrez — Autonomous Networks Research Group (ANRG), University of Southern California