netflowsim

Active Development

← Back to Projects

The Concept

netflowsim provides rapid, massive-scale network performance analysis by using analytic queuing models and Monte Carlo simulations instead of packet-level discrete event simulation. It enables network engineers to validate topologies and routing strategies against billions of flow iterations in seconds.

Quick Facts

What This Is

High-performance flow-based network simulator that uses analytic queuing models and Monte Carlo simulations for rapid topology and routing validation at massive scale.

Why Flow-Based Simulation

Traditional approaches force a choice between:

• Packet-level simulation (high fidelity, slow, small scale)
• Algorithmic analysis (fast, limited accuracy)

netflowsim bridges this gap by modeling flows through queuing theory, achieving:

• Speed: 1M+ flows/second on multi-core hardware
• Scale: 10k+ node topologies
• Accuracy: Validated against theoretical benchmarks

Key Features

Phase 1: Foundation (✅ Complete)

• High-Performance Engine: Rust + petgraph for graph operations
• Parallel Processing: rayon for multi-core utilization
• Queuing Models:
  • M/M/1 (Markov arrivals, Markov service)
  • M/D/1 (Markov arrivals, Deterministic service)
  • M/G/1 (Markov arrivals, General service)
• Monte Carlo Simulation: Billions of flow iterations in seconds
• GeoJSON Export: Congestion hotspot visualization
• Statistics: Mean latency, link utilization, throughput

Phase 2: Routing & Path Tracing (✅ Complete)

• Path Tracing Engine: Calculate end-to-end paths from FIB data
• Routing Matrix Generation: Convert FIB tables to flow routing decisions
• ECMP Support: Equal-cost multi-path with recursive resolution
• Cycle Detection: Per-path loop detection for safe path exploration
• FIB Ingestion: Import forwarding tables from external sources
• Interface-to-Link Mapping: Automatic correlation of FIB interfaces with topology edges

Phase 3: Simulator Interop (🔄 In Progress)

• Topology ingestion from topogener
• Bi-directional integration with packet-level network simulator
• Routing matrix exchange format

Phase 4: Advanced Analysis (Planned)

• CDF plots for latency and throughput distributions
• Automated bottleneck identification
• Performance regression detection

Phase 5: Dynamic Networks (Planned)

• Runtime topology changes (link failures, node additions)
• Dynamic routing matrix re-evaluation
• Failure scenario modeling

Example Usage

# netflowsim
$ netflowsim simulate \
    --topology topology.json \
    --traffic traffic.json \
    --routing routing.json \
    --output results.json

Loaded topology: 128 nodes, 256 links
Loaded traffic: 10000 flows
Running Monte Carlo simulation (1M iterations)...
Completed in 1.2s

Results written to results.json
GeoJSON visualization: network.geojson

# netflowsim
$ netflowsim generate-routing \
    --topology topology.json \
    --fibs examples/fibs/ \
    --output routing.json

Loaded topology: 128 nodes
Ingested FIBs for 128 routers
Traced 16384 source-destination pairs
ECMP paths found: 2841
Routing matrix written to routing.json

Performance Metrics

• Throughput: 1M+ flows/second (baseline)
• Scale: Tested with 10k+ node topologies
• Latency: Sub-second results for typical network sizes
• Accuracy: Validated against M/M/1 and M/D/1 theoretical values

Tech Stack

• Language: Rust
• Graph Library: Petgraph
• Parallelism: Rayon (multi-core processing)
• Serialization: Serde (JSON), GeoJSON
• Visualization: Martin (Tileserver), MVT (Mapbox Vector Tiles)

Use Cases

• Topology Validation: Test network designs before deployment
• Routing Strategy Comparison: Evaluate different routing algorithms
• Capacity Planning: Identify congestion points under load
• What-If Analysis: Model link failures and capacity changes
• Integration Testing: Validate packet simulator routing logic

← Back to Projects