ASNPrefixMap

High-performance C++20 tool for building ASN → Prefix mappings from live BGP streams (RIS Live).

Builds a near-complete global routing view in ~30–40 minutes using real-time data.

---

Dataset Convergence

The system ingests live BGP updates and tracks how quickly new routing information is discovered.

• Rapid growth phase as new data arrives
• Gradual slowdown
• Plateau when most globally visible routes are observed

Plateau detection is automatic and based on rolling averages of prefix discovery rate.

---

Example Output

[plateau] uptime_sec=2163.51 uptime_hms=00:36:03
[plateau] active_prefixes=1293782 total_unique_asns_ever_seen=84951
[plateau] note=table appears broadly complete; you may stop if a stable mapping is enough

raw_messages_received=26661999
parsed_events_total=69416421
announces_applied=62868720
withdraws_applied=6547701

runtime_sec=5697.49
runtime_hms=01:34:57
plateau_detected=true
plateau_uptime_sec=2163.51
plateau_uptime_hms=00:36:03

Key Features

Real-time ingestion via RIS Live WebSocket

Efficient in-memory routing state using binary prefixes and PeerId

Snapshot persistence in a human-readable format

Plateau detection for dataset convergence

Automatic reconnect on transient network failures

Growth statistics for ASN and prefix discovery rates

Architecture

RIS Live WebSocket
        ↓
    JSON Parser
        ↓
     BgpEvent
        ↓
  PeerRegistry (PeerId)
        ↓
 RoutingState (binary prefixes)
        ↓
 Snapshot / Export

Performance Design

The system is designed as a streaming pipeline with a focus on minimizing hot-path overhead.

Key optimizations:

• Incremental prefix-origin selection
  (avoids full recomputation on every update)

• Packed observation storage
  (swap-with-back removal instead of erase)

• No temporary event containers
  (parser emits events directly via callback)

• No temporary event objects
  (fields passed directly instead of materializing structures)

• PeerId cached per message
  (avoids repeated hash/map lookups)

• Single-threaded stats sampling
  (no locking in the ingest path)

• Prefix strings passed by reference from JSON
  (avoids per-event string allocations)

Remaining costs / trade-offs:

• JSON is parsed via DOM (nlohmann::json) for simplicity
• PeerInfo stores owning strings (for snapshot/export stability)
• Prefix parsing still happens per event

The current design prioritizes clarity and correctness, while incrementally reducing hot-path overhead.

Plateau Detection

Plateau is detected using a rolling average of prefix discovery rate.

When the rate drops below a threshold for a sustained period, the system reports that the routing table is broadly complete.

Plateau detection is heuristic and does not guarantee full completeness.

Usage

mkdir build
cd build
cmake ..
make
./asn_prefix_map

Future Work

Query tool for snapshot

HTTP API

Docker image