NetPilot

Defense & Government Research Lab

Reproducible MANET and mesh protocol experiments in the cloud.

AI-built multi-vendor mesh topologies for protocol research under degraded conditions. Cloud-hosted. Artifact-ready for publication. Complementary to CORE/EMANE for RF-channel fidelity research, with cloud-native iteration speed for protocol-layer work.

Babel / OSPF / IS-IS comparison
tc netem impairment injection
Artifact-ready reproducibility

Mesh and MANET research tools compared

Honest positioning. NetPilot is not a replacement for CORE/EMANE on RF-channel fidelity (that is the established open-source reference environment). NetPilot is the cloud iteration layer for protocol-level research, multi-vendor comparison, and artifact-ready reproducibility.

DimensionCORE/EMANEns-3ContainerLabEVE-NGGNS3NetPilot
AI-built from plain-English prompt❌❌❌❌❌✅
Cloud self-serve❌ on-prem❌ self-hosted❌ self-hosted❌ self-hosted VM❌ self-hosted✅
Real NOS CLIs⚠️ via VMs❌ simulator✅ BYOI✅ BYOI✅ BYOI✅ FRR + commercial
RF-channel fidelity (PHY modeling)✅ reference-grade⚠️ modules❌ out of scope❌ out of scope❌ out of scope❌ out of scope
Artifact-ready reproducibility⚠️ scenario files✅ scripts⚠️ YAML topo⚠️ lab files⚠️ project files✅ prompt = artifact
Multi-vendor NOS interop⚠️ via VMs❌✅ BYOI✅ BYOI✅ BYOI✅
Setup timeDays (install + scenario)Days (scripts)Hours (install + images)Hours–days (VM + images)Hours (install + images)~2 minutes
Anchor experiment

Babel vs OSPF vs IS-IS under progressive packet loss

A canonical mesh-routing experiment: compare three routing protocols on identical topologies under increasing packet-loss conditions. Which converges fastest? Which is most stable? Which misbehaves at what loss threshold?

The NetPilot workflow: describe the experiment in plain English — "8-node mesh with FRR running Babel, 8-node mesh with FRR running OSPF, and 8-node mesh with FRR running IS-IS — same topology, same node positions. tc netem injects correlated packet loss progressively from 0% to 20%. Capture per-protocol convergence time, route table stability, and adjacency flaps."

NetPilot generates FRR configurations for all three protocols, deploys all three labs in ~2 minutes each, runs the impairment script, and produces per-protocol data that's reproducible by any researcher with the same prompt. The prompt + generated configs become the artifact.

→ AI-powered MANET research labs→ OSPF vs Babel under link failure

Complement to CORE/EMANE — not replacement

CORE/EMANE wins

  • • RF-channel fidelity and waveform modeling
  • • PHY-layer channel effects and interference
  • • Established open-source reference lineage
  • • On-premises deployment for regulated environments

NetPilot wins

  • • AI-built topologies from plain-English prompts
  • • Cloud self-serve — minutes to running lab
  • • Multi-vendor NOS (FRR + commercial) for interop research
  • • Artifact-ready reproducibility for publication

Use both. CORE/EMANE for RF-fidelity validation; NetPilot for protocol-layer iteration and multi-vendor comparison. The two target different layers of the stack.

Use cases for defense and government research teams

Four research workflows where cloud-native protocol iteration adds speed to an existing RF-fidelity toolchain.

Cross-protocol mesh behavior

Compare Babel vs OSPF vs IS-IS convergence and route stability under packet loss on identical mesh topologies. Reproducible per-protocol behavioral data in the cloud. The comparison cited in the NetPilot OSPF-vs-Babel analysis is the canonical starting point.

OSPF vs Babel under link failure →

Packet loss, latency, and jitter impairment research

tc netem on Linux endpoints provides scriptable, reproducible IP-layer impairment — uniform and correlated packet loss, latency, jitter, duplication, reordering, rate limiting, and link-flap patterns. Scripted, repeatable, capturable per experiment.

Multi-vendor mesh research

FRR mesh interoperating with commercial NOSes — real CLIs, real protocol behavior, real adjacency state. Useful for researching cross-implementation behavior that single-vendor simulators cannot surface.

AI-powered MANET research labs →

Reproducible artifacts for FFRDC / academic publication

The prompt + generated configs + deployable topology form an artifact another researcher can run in minutes. Matches SIGCOMM / CoNEXT / IMC artifact-evaluation expectations — addresses the environment-setup bottleneck that drives reproducibility rates below one-third in networking research.

Protocols and impairments supported

IP and routing-protocol layer coverage. For RF-channel fidelity, pair with CORE/EMANE.

  • Routing: BGP, OSPF (multi-area), IS-IS (multi-level, wide metrics), Babel, RIP, EIGRP (on vendor NOS)
  • EVPN (Type-2/3/5, symmetric/asymmetric IRB) and VXLAN
  • SRv6 (uSID endpoint behaviors, L3VPN over SRv6) and SR-MPLS
  • PIM (ASM, SSM) for multicast mesh research
  • BFD (multi-hop, authenticated)
  • Impairments (tc netem): packet loss (uniform, correlated), latency, jitter, duplication, reordering, rate limit, link flap
  • Malformed packet injection via Scapy (protocol-layer only)
  • Multi-vendor NOS: FRR + Cisco IOL / Juniper cRPD / Arista cEOS / Nokia SR Linux (BYOI)

Scope note

NetPilot operates at the IP and routing-protocol layer. Research workflows requiring RF-channel fidelity (PHY modeling, waveform simulation) are best served by CORE/EMANE. NetPilot does not hold FedRAMP, IL4, or IL5 authorization; research labs with those compliance requirements should use on-premises tooling.

Defense & Government Research FAQ

Scenario-phrased questions from research practitioners.

Describe the experiment in plain English — for example, '8-node mesh with FRR running Babel on one topology, FRR running OSPF on a second identical topology, and FRR running IS-IS on a third. tc netem injects 0% to 20% correlated packet loss progressively. Capture convergence times and route stability per protocol.' NetPilot deploys all three in ~2 minutes, runs the impairment script, and produces reproducible per-protocol data.
No. CORE/EMANE is the open-source reference environment for RF-channel fidelity modeling and waveform research. NetPilot is complementary — it targets protocol-layer iteration, multi-vendor NOS behavior, and reproducible artifacts. Use CORE/EMANE for RF-layer fidelity; use NetPilot for control-plane iteration, multi-vendor protocol comparison, and publication-ready reproducibility.
Yes. NetPilot's prompt + generated configs + deployable topology form a reusable artifact that another researcher can run in minutes. This matches SIGCOMM / CoNEXT / IMC artifact-evaluation program expectations. Reproducibility in networking research has historically been low — around 32% by one ACM survey — and the prompt-as-artifact pattern directly addresses the environment-setup bottleneck that drives that number.
NetPilot's Linux endpoints use tc netem for scriptable, reproducible impairment: packet loss (uniform and correlated), latency, jitter, duplication, reordering, rate limiting, and link-flap patterns. Scapy is available for malformed or custom packet injection at the routing-protocol layer. The impairments are IP-layer — CORE/EMANE remains the correct tool for RF-channel fidelity modeling.
No — those are out of scope. NetPilot operates at the IP and routing-protocol layer. For waveform simulation, PHY-layer interference modeling, or RF-channel fidelity, CORE/EMANE remains the correct tool. NetPilot's wedge is protocol-layer iteration speed, multi-vendor comparison, and artifact-ready reproducibility.
Yes. NetPilot's default FRR image runs Babel alongside BGP, OSPF, IS-IS, EVPN, and SRv6. Describe the mesh topology, prompt the experiment, and SSH into any node to verify Babel state. The FRR cloud lab guide walks through the six-protocol FRR stack with copy-pasteable prompts.
ContainerLab, EVE-NG, and GNS3 are the canonical self-hosted multi-vendor lab platforms — powerful but self-operated. You provision the host, source the vendor images, and maintain the install. NetPilot is the cloud-hosted and AI-built equivalent: describe the topology in plain English and get a running lab in ~2 minutes, with the same real multi-vendor CLIs. Use the self-hosted tools when on-premises deployment is required; use NetPilot when iteration speed and artifact reproducibility matter more than local infrastructure control.
No. NetPilot is a commercial cloud platform and does not hold FedRAMP, IL4, or IL5 authorization. Research labs with those compliance requirements should use on-premises tooling (CORE/EMANE or self-hosted ContainerLab / EVE-NG / GNS3). NetPilot is for research teams whose workflow does not require those authorizations.

Related reading

Guide

AI-Powered MANET Research Labs

Reproducible mesh network experiments in the cloud — for defense, academic, and community mesh research.

Tutorial

OSPF vs Babel Under Link Failure

Head-to-head routing protocol convergence comparison under impairment.

Tutorial

FRRouting Cloud Labs

Six FRR protocols in one prompt — BGP, OSPF, IS-IS, Babel, EVPN, SRv6.

Guide

BGP Convergence Research

Methodology and artifact patterns for BGP convergence research.

Comparison

Keysight vs VIAVI vs NetPilot: Research Lab Comparison

Honest comparison across research tooling categories.

Hub

Network Research Lab

The parent hub — all six research segments.

Ready to iterate on mesh protocols in minutes?

Dedicated environments, custom vendor image support, workflow integration — talk to us about a research plan. Or spin up a free lab and try a Babel vs OSPF comparison yourself.