arxiv: 2605.01094 · v1 · submitted 2026-05-01 · 💻 cs.DC · cs.NI

Recognition: unknown

ncsim: A Lightweight Simulator for Networked Edge Computing with Wireless Interference Modeling

Bhaskar Krishnamachari , Maya Gutierrez , Jared Coleman

Authors on Pith no claims yet

Pith reviewed 2026-05-09 18:02 UTC · model grok-4.3

classification 💻 cs.DC cs.NI

keywords edge computingwireless interferenceDAG schedulingsimulatorrank inversionmakespanCSMA/CA

0 comments

The pith

Ignoring wireless interference when choosing edge computing schedulers can select the worst algorithm in more than a quarter of cases.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows that schedulers for tasks on wireless edge networks must account for interference to avoid poor choices. Without it, the scheduler that looks best on paper can actually take up to 2.7 times longer to complete the work than a basic round-robin approach. The authors built ncsim to simulate both the task dependencies and the wireless contention in one tool. Their tests across many scenarios found that this mismatch occurs often enough to change which algorithm should be used.

Core claim

ncsim demonstrates that rank inversions occur when wireless interference is modeled, causing the interference-free optimal scheduler to produce makespans up to 2.7 times worse than a simple baseline, with inversions in 27.8% of tested scenarios and 50% in larger 100-node graphs.

What carries the argument

ncsim, a lightweight discrete-event simulator that integrates DAG workflow scheduling with IEEE 802.11 CSMA/CA interference modeling in a single package.

If this is right

Schedulers must be evaluated with interference included to produce correct performance rankings.
In many scenarios the scheduler optimal without interference yields makespans up to 2.7 times longer than round-robin.
Inversion rates increase with network scale, reaching 50 percent on 100-node random geometric graphs.
A joint simulator is required because isolated compute and communication models produce misleading algorithm selections.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Application developers may need to re-test previously chosen schedulers once interference is added to the evaluation.
The same mismatch between isolated and joint models could appear in other networked resource allocation problems.
Running the same factorial experiments on physical hardware would directly test whether the simulated inversion rates persist.

Load-bearing premise

The IEEE 802.11 CSMA/CA interference model accurately captures real-world wireless conditions so that the observed rank inversions would hold on actual hardware.

What would settle it

Deploying the compared schedulers on a real wireless edge computing testbed with 802.11 devices and measuring if the makespan differences and inversions match the simulation results.

Figures

Figures reproduced from arXiv: 2605.01094 by Bhaskar Krishnamachari, Jared Coleman, Maya Gutierrez.

**Figure 1.** Figure 1: High-level architecture of ncsim. interference modeling view at source ↗

**Figure 2.** Figure 2: Experiment 1: PHY data rate vs. link distance showing view at source ↗

**Figure 4.** Figure 4: Experiment 4 topology: three parallel 30 m links at view at source ↗

**Figure 5.** Figure 5: Reproduction of Bianchi [9] view at source ↗

**Figure 6.** Figure 6: Cross-validation of ncsim against ns-3. (a) n-way saturated goodput; (b) two-link throughput vs. separation. ncsim tracks ns-3 within 0.3–11% across all regimes. Parameters per Table IX; each ns-3 point: mean ± 95% CI over 20 seeds. predictions differ by only 0.3% (3.78 vs. 3.77 MB/s). The maximum error of 7.6% occurs at n=7, where the Bianchi fixedpoint approximation slightly overestimates MAC efficienc… view at source ↗

**Figure 7.** Figure 7: Link utilization under widest-path (left) vs. shortest-path view at source ↗

**Figure 8.** Figure 8: Scheduling regret ratios. Each bar shows the ratio view at source ↗

**Figure 9.** Figure 9: Interference slowdown vs. CCR (3×3 grid, HEFT, shortest-path). Slowdown peaks at intermediate CCR values where transfers are long enough to contend but not so large that the scheduler collapses tasks onto one node. The 20-task DAG reaches 7.3× at CCR≈ 0.53. a CCR of approximately 0.5 (where transfer time and compute time are comparable). 2) Multi-DAG Contention: In practice, edge computing platforms rarely… view at source ↗

**Figure 10.** Figure 10: Multi-DAG makespan scaling on 3×3 grid (5-task DAGs, HEFT, shortest-path). Contention amplifies superlinearly: the gap between the curves widens from 2.26× (k=1) to 4.08× (k=5). between the two curves widens with each additional DAG: the slowdown factor grows from 2.26× at k=1 to 2.98× at k=2 to 4.08× at k=5, indicating a positive feedback loop. Each additional DAG contributes its own transfers to the co… view at source ↗

read the original abstract

Evaluating DAG task schedulers for wireless edge computing requires jointly modeling compute placement and wireless interference, yet existing tools treat them in isolation. This gap leads to rank inversions: the scheduler that appears optimal under an interference-free model can be the worst choice under realistic wireless conditions. We present ncsim, a lightweight discrete-event simulator that bridges this gap by combining DAG workflow scheduling with physically-grounded IEEE 802.11 CSMA/CA interference modeling in a single Python package. A 108-run factorial experiment reveals rank inversions in 27.8% of scenarios, with the interference-free-optimal scheduler producing up to 2.7x worse makespan than a simple round-robin baseline; scaling to a 100-node random geometric graph raises the inversion rate to 50%. These rank inversions show that interference-free evaluation can select the wrong algorithm entirely, justifying the design and use of ncsim.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

ncsim combines DAG scheduling and CSMA/CA interference in one open Python tool and reports rank inversions in 27.8% of cases, but the inversion rates rest on an unvalidated wireless model.

read the letter

The paper's core point is straightforward: schedulers that look best when you ignore wireless interference can turn out much worse once you add realistic CSMA/CA effects, and ncsim is the tool that makes it easy to see this in the same run. It packages a discrete-event simulator that handles both task graphs and 802.11 backoff, collisions, and SINR reception, then runs a 108-scenario factorial design plus a 100-node scaling case to quantify how often the ordering flips (27.8% and 50% respectively) and how large the makespan penalty can get (up to 2.7x). That combination and the concrete inversion counts are new relative to prior separate tools for scheduling or wireless simulation alone. The code is lightweight and open, which lowers the barrier for anyone who wants to test their own scheduler under interference without stitching multiple packages together. The authors also keep the focus narrow on edge-computing DAGs, so the experiments stay manageable. The main limitation is the lack of any anchor for the interference model itself. There are no calibration runs against hardware testbeds, no packet-delivery or latency comparisons to ns-3 or analytical CSMA models, and no sensitivity checks on carrier-sensing range or propagation assumptions. Without those, the reported inversion rates and 2.7x gaps could be artifacts of the specific implementation choices rather than general wireless behavior. The abstract gives the headline percentages but the full text would need to show the exact parameter settings and any statistical checks to let readers judge how robust the 108-run design really is. This work is aimed at researchers who already build or evaluate schedulers for wireless edge systems and who have felt the gap between compute-only and network-aware evaluation. A reader who needs a ready-to-use simulator for quick what-if experiments on small-to-medium networks will find it immediately useful. It is coherent on its own terms and engages the right prior literature on both scheduling and wireless modeling, so it deserves a serious referee. I would send it out, but the review should focus on adding at least one external validation step for the CSMA/CA component before acceptance.

Referee Report

2 major / 1 minor

Summary. The paper introduces ncsim, a lightweight Python discrete-event simulator that integrates DAG task scheduling with physically-grounded IEEE 802.11 CSMA/CA wireless interference modeling (backoff, collisions, SINR-based reception) for networked edge computing. It claims that a 108-run factorial experiment demonstrates rank inversions in 27.8% of scenarios, where the interference-free-optimal scheduler yields up to 2.7x worse makespan than round-robin, with the inversion rate rising to 50% on a 100-node random geometric graph; this is used to argue that interference-free evaluation can select entirely wrong algorithms and to justify the simulator.

Significance. If the interference model holds, the result is significant: it identifies a concrete failure mode in current scheduler evaluation practice for wireless edge systems and shows that joint compute-wireless modeling can reverse algorithm rankings. The lightweight, open Python implementation is a practical strength that could aid adoption and reproducibility in the distributed computing community.

major comments (2)

The central empirical claim (rank inversions in 27.8% of the 108-run factorial design, up to 2.7x makespan gap) rests on the accuracy of ncsim's IEEE 802.11 CSMA/CA implementation, yet the manuscript reports no calibration, no packet-delivery-ratio or latency comparison against hardware testbeds, and no cross-validation against ns-3 or analytical CSMA models. This is load-bearing for the inference that the observed inversions would persist in real wireless conditions.
No statistical significance testing, confidence intervals, or sensitivity analysis on parameter choices (e.g., carrier-sensing range, propagation model, synchronization assumptions) is described for the reported percentages or the scaling result on the 100-node graph; without these, it is unclear whether the 27.8% and 50% inversion rates are robust or artifacts of specific simulator settings.

minor comments (1)

The abstract and results would benefit from an explicit table or paragraph listing the exact parameter ranges and factor levels used in the 108-run experiment.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and insightful comments. They correctly identify areas where additional validation and statistical analysis would strengthen the empirical claims. We address each major comment point by point below, outlining the specific revisions we will make to the manuscript.

read point-by-point responses

Referee: The central empirical claim (rank inversions in 27.8% of the 108-run factorial design, up to 2.7x makespan gap) rests on the accuracy of ncsim's IEEE 802.11 CSMA/CA implementation, yet the manuscript reports no calibration, no packet-delivery-ratio or latency comparison against hardware testbeds, and no cross-validation against ns-3 or analytical CSMA models. This is load-bearing for the inference that the observed inversions would persist in real wireless conditions.

Authors: We acknowledge that the manuscript does not currently include calibration against hardware or explicit cross-validation of the wireless model. The ncsim CSMA/CA implementation is based directly on the IEEE 802.11 standard specifications for exponential backoff, collision detection, and SINR-threshold reception using standard log-distance propagation. To address the concern, the revised manuscript will add a new validation subsection. This will report packet delivery ratio and end-to-end latency results from ncsim compared against both closed-form analytical CSMA/CA models and equivalent ns-3 simulations under identical topologies and traffic patterns. We will also explicitly state that hardware testbed calibration remains future work, as it requires physical infrastructure not available for this study, while noting that the current results demonstrate rank inversions under a standard, physically grounded model rather than claiming direct real-world equivalence. revision: yes
Referee: No statistical significance testing, confidence intervals, or sensitivity analysis on parameter choices (e.g., carrier-sensing range, propagation model, synchronization assumptions) is described for the reported percentages or the scaling result on the 100-node graph; without these, it is unclear whether the 27.8% and 50% inversion rates are robust or artifacts of specific simulator settings.

Authors: We agree that the absence of statistical analysis and sensitivity testing leaves the robustness of the 27.8% and 50% inversion rates open to question. In the revised manuscript we will add bootstrap 95% confidence intervals for both inversion percentages. We will also include a sensitivity analysis section that systematically varies carrier-sensing range, path-loss exponent, and node synchronization assumptions across the factorial design and the 100-node random geometric graph. Statistical tests (chi-squared for inversion rates and paired t-tests for makespan differences) will be reported to establish significance. These additions will demonstrate that the observed rank inversions are not artifacts of the chosen default parameters. revision: yes

Circularity Check

0 steps flagged

No circularity: results are direct empirical outputs from simulator runs

full rationale

The paper introduces ncsim as a new discrete-event simulator combining DAG scheduling with IEEE 802.11 CSMA/CA modeling, then reports outcomes from a 108-run factorial experiment and a 100-node scaling test. These produce the observed rank inversion rates (27.8% and 50%) and makespan ratios (up to 2.7x) as direct simulation results on generated scenarios. No equations, parameter fits, predictions, or self-citations appear in the text that would reduce any claim to its own inputs by construction. The central findings are therefore self-contained empirical observations rather than derivations that collapse to fitted values or prior author work.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The simulator rests on standard discrete-event simulation mechanics and the IEEE 802.11 CSMA/CA backoff model; no new physical constants or entities are introduced. The 108-run experiment implicitly assumes that the chosen task graphs, network topologies, and scheduler implementations are representative.

axioms (2)

domain assumption IEEE 802.11 CSMA/CA interference model accurately captures real wireless contention
Invoked when the paper states 'physically-grounded IEEE 802.11 CSMA/CA interference modeling'
standard math Discrete-event simulation faithfully reproduces scheduler behavior under the modeled interference
Core of any discrete-event simulator

pith-pipeline@v0.9.0 · 5460 in / 1411 out tokens · 35490 ms · 2026-05-09T18:02:54.796798+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

34 extracted references · 6 canonical work pages

[1]

Edge Computing: Vision and Challenges,

W. Shi, J. Cao, Q. Zhang, Y . Li, and L. Xu, “Edge Computing: Vision and Challenges,”IEEE Internet of Things Journal, vol. 3, no. 5, pp. 637–646, Oct. 2016

2016
[2]

ns-3: A Discrete-Event Network Simulator,

ns-3 Consortium, “ns-3: A Discrete-Event Network Simulator,” https: //www.nsnam.org/, accessed Mar. 2026

2026
[3]

An Overview of the OMNeT++ Simulation Environment,

A. Varga and R. Hornig, “An Overview of the OMNeT++ Simulation Environment,” inProc. ICST SIMUTools, 2008

2008
[4]

Versatile, Scalable, and Accurate Simulation of Distributed Applications and Platforms,

H. Casanova, A. Giersch, A. Legrand, M. Quinson, and F. Suter, “Versatile, Scalable, and Accurate Simulation of Distributed Applications and Platforms,”Journal of Parallel and Distributed Computing, vol. 74, no. 10, pp. 2899–2917, 2014

2014
[5]

WRENCH: A Framework for Simulating Workflow Management Systems,

H. Casanova, S. Pandey, J. Oeth, R. Tanaka, F. Suter, and R. Fer- reira da Silva, “WRENCH: A Framework for Simulating Workflow Management Systems,” inProc. WORKS Workshop, 2018, pp. 74–85

2018
[6]

CloudSim: A Toolkit for Modeling and Simulation of Cloud Computing Environments and Evaluation of Resource Provisioning Algorithms,

R. N. Calheiros, R. Ranjan, A. Beloglazov, C. A. F. De Rose, and R. Buyya, “CloudSim: A Toolkit for Modeling and Simulation of Cloud Computing Environments and Evaluation of Resource Provisioning Algorithms,”Software: Practice and Experience, vol. 41, no. 1, pp. 23–50, 2011

2011
[7]

iFogSim: A Toolkit for Modeling and Simulation of Resource Management Tech- niques in the Internet of Things, Edge and Fog Computing Environments,

H. Gupta, A. Vahid Dastjerdi, S. K. Ghosh, and R. Buyya, “iFogSim: A Toolkit for Modeling and Simulation of Resource Management Tech- niques in the Internet of Things, Edge and Fog Computing Environments,” Software: Practice and Experience, vol. 47, no. 9, pp. 1275–1296, 2017

2017
[8]

EdgeCloudSim: An Environment for Performance Evaluation of Edge Computing Systems,

C. Sonmez, A. Ozgovde, and C. Ersoy, “EdgeCloudSim: An Environment for Performance Evaluation of Edge Computing Systems,”Transactions on Emerging Telecommunications Technologies, vol. 29, no. 11, e3493, 2018

2018
[9]

Performance Analysis of the IEEE 802.11 Distributed Coor- dination Function,

G. Bianchi, “Performance Analysis of the IEEE 802.11 Distributed Coor- dination Function,”IEEE Journal on Selected Areas in Communications, vol. 18, no. 3, pp. 535–547, Mar. 2000

2000
[10]

Performance-Effective and Low-Complexity Task Scheduling for Heterogeneous Computing,

H. Topcuoglu, S. Hariri, and M.-Y . Wu, “Performance-Effective and Low-Complexity Task Scheduling for Heterogeneous Computing,”IEEE Trans. Parallel Distrib. Syst., vol. 13, no. 3, pp. 260–274, 2002

2002
[11]

A Comparison of List Schedules for Parallel Processing Systems,

T. L. Adam, K. M. Chandy, and J. R. Dickson, “A Comparison of List Schedules for Parallel Processing Systems,”Communications of the ACM, vol. 17, no. 12, pp. 685–690, 1974

1974
[12]

Static Scheduling Algorithms for Allocating Directed Task Graphs to Multiprocessors,

Y .-K. Kwok and I. Ahmad, “Static Scheduling Algorithms for Allocating Directed Task Graphs to Multiprocessors,”ACM Computing Surveys, vol. 31, no. 4, pp. 406–471, 1999

1999
[13]

A Compile-Time Scheduling Heuristic for Interconnection-Constrained Heterogeneous Processor Architectures,

G. C. Sih and E. A. Lee, “A Compile-Time Scheduling Heuristic for Interconnection-Constrained Heterogeneous Processor Architectures,” IEEE Trans. Parallel Distrib. Syst., vol. 4, no. 2, pp. 175–187, 1993

1993
[14]

Comparing Task Graph Scheduling Algorithms: An Adversarial Approach,

J. Coleman and B. Krishnamachari, “Comparing Task Graph Scheduling Algorithms: An Adversarial Approach,” arXiv preprint arXiv:2403.07120, 2024

work page arXiv 2024
[15]

SAGA: Scheduling Algorithms Gathered (software),

J. Coleman, B. Krishnamachari, and contributors, “SAGA: Scheduling Algorithms Gathered (software),” https://github.com/anrgusc/saga, 2024

2024
[16]

Parame- terized Task Graph Scheduling Algorithm for Comparing Algorithmic Components,

J. Coleman, R. V . Agrawal, E. Hirani, and B. Krishnamachari, “Parame- terized Task Graph Scheduling Algorithm for Comparing Algorithmic Components,” arXiv preprint arXiv:2403.07112, 2024

work page arXiv 2024
[17]

Jupiter: A Networked Computing Architecture,

P. Ghosh, Q. Nguyen, P. K. Sakulkar, J. A. Tran, A. Knezevic, J. Wang, Z. Lin, B. Krishnamachari, M. Annavaram, and S. Avestimehr, “Jupiter: A Networked Computing Architecture,” inProc. UCC Companion, 2021

2021
[18]

Tactical Jupiter: Dynamic Scheduling of Dispersed Computations in Tactical MANETs,

A. Poylisher, A. Cichocki, K. Guo, J. Hunziker, L. Kant, B. Kr- ishnamachari, S. Avestimehr, and M. Annavaram, “Tactical Jupiter: Dynamic Scheduling of Dispersed Computations in Tactical MANETs,” inMILCOM 2021, 2021

2021
[19]

Design and Experimental Evaluation of Algorithms for Optimizing the Throughput of Dispersed Computing,

X. Zhao, D. Hu, and B. Krishnamachari, “Design and Experimental Evaluation of Algorithms for Optimizing the Throughput of Dispersed Computing,” arXiv preprint arXiv:2112.13875, 2021

work page arXiv 2021
[20]

I-BOT: Interference-Based Orchestration of Tasks for Dynamic Unman- aged Edge Computing,

S. Suryavansh, C. Bothra, K. T. Kim, M. Chiang, C. Peng, and S. Bagchi, “I-BOT: Interference-Based Orchestration of Tasks for Dynamic Unman- aged Edge Computing,” arXiv preprint arXiv:2011.05925, 2020

work page arXiv 2011
[21]

DAG-based Task Orchestration for Edge Computing,

X. Li, M. Abdallah, S. Suryavansh, M. Chiang, K. T. Kim, and S. Bagchi, “DAG-based Task Orchestration for Edge Computing,” inProc. SRDS, 2022, pp. 23–34

2022
[22]

Dynamic DAG-Application Scheduling for Multi-Tier Edge Computing in Heterogeneous Networks,

X. Li, M. Abdallah, Y .-Y . Lou, M. Chiang, K. T. Kim, and S. Bagchi, “Dynamic DAG-Application Scheduling for Multi-Tier Edge Computing in Heterogeneous Networks,” arXiv preprint arXiv:2409.10839, 2024

work page arXiv 2024
[23]

LA VEA: Latency- aware Video Analytics on Edge Computing Platform,

S. Yi, Z. Hao, Q. Zhang, Q. Zhang, W. Shi, and Q. Li, “LA VEA: Latency- aware Video Analytics on Edge Computing Platform,” inProc. SEC, 2017

2017
[24]

Hetero- Edge: Orchestration of Real-time Vision Applications on Heterogeneous Edge Clouds,

W. Zhang, S. Li, L. Liu, Z. Jia, Y . Zhang, and D. Raychaudhuri, “Hetero- Edge: Orchestration of Real-time Vision Applications on Heterogeneous Edge Clouds,” inIEEE INFOCOM, 2019, pp. 1270–1278

2019
[25]

Distributed and Application-aware Task Scheduling in Edge-clouds,

L. Lin, P. Li, J. Xiong, and M. Lin, “Distributed and Application-aware Task Scheduling in Edge-clouds,” arXiv preprint arXiv:1902.04362, 2019

work page arXiv 1902
[26]

Joint offloading and scheduling deci- sions for DAG applications in mobile edge computing,

J. Liang, K. Li, C. Liu, and K. Li, “Joint offloading and scheduling deci- sions for DAG applications in mobile edge computing,”Neurocomputing, vol. 424, pp. 160–171, 2021

2021
[27]

Energy-Constrained DAG Scheduling on Edge and Cloud Servers with Overlapped Communication and Computation,

K. Li, “Energy-Constrained DAG Scheduling on Edge and Cloud Servers with Overlapped Communication and Computation,”Journal of Grid Computing, 2024

2024
[28]

SecDS: A security-aware DAG task scheduling strategy for edge computing,

L. Long, Z. Liu, J. Shen, and Y . Jiang, “SecDS: A security-aware DAG task scheduling strategy for edge computing,”Future Generation Computer Systems, p. 107627, 2025

2025
[29]

Fog and IoT: An Overview of Research Opportunities,

M. Chiang and T. Zhang, “Fog and IoT: An Overview of Research Opportunities,”IEEE Internet of Things Journal, vol. 3, no. 6, pp. 854– 864, 2016

2016
[30]

Clarifying Fog Computing and Networking: 10 Questions and Answers,

M. Chiang, S. Ha, I Chih-Lin, F. Risso, and T. Zhang, “Clarifying Fog Computing and Networking: 10 Questions and Answers,”IEEE Communications Magazine, vol. 55, no. 4, pp. 18–20, 2017

2017
[31]

Coded Edge Computing,

K. T. Kim, C. Joe-Wong, and M. Chiang, “Coded Edge Computing,” in IEEE INFOCOM, 2020, pp. 237–246

2020
[32]

Saturation Throughput Analysis of IEEE 802.11 in the Presence of Non Ideal Transmission Channel and Capture Effects,

F. Daneshgaran, M. Laddomada, F. Mesiti, M. Mondin, and M. Zanolo, “Saturation Throughput Analysis of IEEE 802.11 in the Presence of Non Ideal Transmission Channel and Capture Effects,”IEEE Trans. Communications, vol. 56, no. 7, pp. 1178–1188, Jul. 2008

2008
[33]

Capture and Retransmission Control in Mobile Radio,

M. Zorzi and R. R. Rao, “Capture and Retransmission Control in Mobile Radio,”IEEE J. Sel. Areas Commun., vol. 12, no. 8, pp. 1289–1298, Oct. 1994

1994
[34]

Capture Effect in IEEE 802.11 Basic Service Area Under Influence of Rayleigh Fading and Near/Far Effect,

Z. Hadzi-Velkov and B. Spasenovski, “Capture Effect in IEEE 802.11 Basic Service Area Under Influence of Rayleigh Fading and Near/Far Effect,” inIEEE PIMRC, 2002

2002