ncsim: A Lightweight Simulator for Networked Edge Computing with Wireless Interference Modeling
A new lightweight simulator shows evaluating schedulers without wireless modeling can pick the worst algorithm 27.8% of the time.
Evaluating DAG task schedulers for wireless edge computing has historically treated compute placement and wireless interference as separate concerns. This gap leads to rank inversions—where the scheduler that appears best under an interference-free model becomes the worst when wireless effects are included. To address this, Bhaskar Krishnamachari, Maya Gutierrez, and Jared Coleman developed ncsim, a lightweight discrete-event simulator written in Python. It bridges both domains by combining DAG workflow scheduling with physically-grounded IEEE 802.11 CSMA/CA interference modeling in a single package. The tool is designed for rapid experimentation and is open-sourced with code and configurations available online.
In a rigorous factorial experiment with 108 runs, ncsim revealed rank inversions in 27.8% of scenarios. The interference-free-optimal scheduler produced up to 2.7x longer makespan than a simple round-robin baseline. When scaled to a 100-node random geometric graph, the inversion rate jumped to 50%, demonstrating that ignoring wireless interference can lead to selecting the wrong algorithm entirely. These findings underscore the necessity of integrated simulation tools like ncsim for anyone building or deploying schedulers in real-world wireless edge environments.
- ncsim jointly models DAG workflow scheduling and IEEE 802.11 CSMA/CA wireless interference in a single Python package.
- Rank inversions occurred in 27.8% of scenarios; the interference-free optimal scheduler was up to 2.7x worse than round-robin.
- Scaling to 100 nodes increased inversion rate to 50%, proving interference-free evaluation can select the wrong algorithm.
Why It Matters
For engineers designing edge computing systems, ncsim prevents costly scheduler mis-selection by modeling real wireless interference.