The Circuit Breaker That Cried Wolf: Measuring BGP Maximum-Prefix Exceedance and Connectivity Impact

This repository contains the public version of the code for our work presented at the 2026 ACM Internet Measurement Conference (IMC '26), October 12–16, 2026, Karlsruhe, Germany.

Abstract. The BGP maximum-prefix limit serves as a network circuit breaker, tearing down sessions when announcements exceed a configured threshold. Despite its operational significance, little is known empirically about how accurately these limits reflect reality, how often they are exceeded, or what connectivity impact results from their exceedance. We present a first large-scale measurement study of BGP maximum-prefix limits in the wild, combining 12 months of RIPE RIS snapshots with PeeringDB metadata and CAIDA AS Rank data. Only 32.49% of RIS-visible ASes appear in PeeringDB, and 21–27% of recent limit fields carry a value of zero, indicating no declared limit; after filtering for recency and non-zero limits, 14,973 ASes form our analytical dataset. Among these, 14.55% (IPv4) and 11.24% (IPv6) exceed their declared limit intermittently, and 2.15% and 2.32% do so in every observed snapshot. These exceedances are driven by organic growth, such as new space acquisition or deaggregation, yet the mechanism responds identically to a route leak: Tier-1 and Major peers account for 18.4% of IPv4 lost sessions despite representing roughly 0.1% of all ASes. Based on our findings, we offer recommendations for operators and IXPs and call for renewed IETF standardization of dynamic limit negotiation.

A deaggregation pushes AS44901 (BelCloud)'s announced prefix count past its PeeringDB maximum-prefix limit (the circuit breaker); peers drop, the operator responds by raising the limit, and peering then recovers. (Figure 1 in the paper.)

Repository layout

.
├── settings.json        # configuration: date range, data paths (relative), collectors, Tier-1 list
├── requirements.txt     # Python dependencies (pinned)
├── notebooks/           # the pipeline + analysis (primary, runnable form)
├── scripts/             # minimal helpers: data fetch + long-running jobs
├── images/              # final figure PDFs from the paper
└── data/                # NOT in git — skeleton only; see each NOTE.md
    ├── raw/             # fetched by the scripts (RIBs, updates, PeeringDB, CAIDA)
    └── processed/       # regenerated by the notebooks (visibility, peers, events, numbers)

Data is not shipped: raw sources are large and re-downloadable, and processed artifacts are regenerated by the code (single source of truth). Every data/ subfolder has a NOTE.md explaining what it holds and which step produces it.

Setup

python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

Requires Python 3.12+. MRT parsing uses bgpkit via pybgpkit-parser.

Configuration

All paths and parameters live in settings.json (paths are relative to the repo root, so the pipeline reads, writes, and saves plots inside the cloned tree). Key fields: START_DATE/END_DATE, DATA_DIR, IMAGE_DIR, N_PROCESSES, the 23 RIS COLLECTORS, and the authoritative TIER1 AS list.

Reproducibility pipeline

Numbered stages run in order. Notebooks are the runnable/readable form; a few heavy stages also ship a script (same number) for long unattended / tmux runs.

#	Step	File(s)	Produces
00	Download RIBs	scripts/00-Download_RIBs.py	raw RIB snapshots
01	Download PeeringDB	scripts/01-Download_PeeringDB.py	raw PeeringDB dumps
02	Parse PeeringDB	notebooks/02-Parse_PeeringDB.ipynb	prefix-limit table
03	Parse RIBs — originated	03-Parse_RIBs_originated (.py/.ipynb)	per-AS originated prefixes
04	Parse RIBs — announced	04-Parse_RIBs_announcement (.py/.ipynb)	per-AS announced prefixes
05–06	Originated high-visibility (+ timeseries)	05-Compute_originated_visibility, 06-Timeseries_originated_visibility	visibility datasets
07–08	Announced high-visibility (+ timeseries)	07-Compute_announced_visibility, 08-Timeseries_announced_visibility	visibility datasets
09–10	Peer graphs + per-AS peer timeseries	09-Compute_Peer_Graphs, 10-Compute_Peer_Timeseries	df_peers
11	PeeringDB coverage & staleness	notebooks/11-PeeringDB_Stats.ipynb	Figs 2–5, Table 3
12	Exceedance analysis	notebooks/12-Exceedance_Analysis.ipynb	Figs 6, 7, 12, Table 4
13	Impact events + lost-peer rank	notebooks/13-Impact_Events.ipynb	Figs 8, 9, Table 5, event sets
14	Visibility-threshold sensitivity	notebooks/14-Sensitivity_Analysis.ipynb	Tables 2, 7
15	Event classification (cause × persistence)	notebooks/15-Event_Classification.ipynb	Table 9
16	Root cause: deaggregation vs. new space	notebooks/16-Root_Cause.ipynb	Fig 11
17–18	Case-study updates: fetch + extract	scripts/17-Fetch_Case_Updates.py, scripts/18-Extract_Case_Updates.py	case replay inputs
19–20	Case-study plots (RIB + 5-min zoom)	notebooks/19-Case_Study_RIB_Plots.ipynb, notebooks/20-Case_Study_Zoom_Plots.ipynb	Fig 1, Fig 10

CAIDA AS Rank (data/raw/AS_rank/asrank-download.py) feeds the rank-based steps (13, 12); scripts/case_studies.py holds the four case definitions imported by steps 17–20.

Reproducing results

• Inspect only — browse the notebooks and the shipped figure PDFs in images/. No compute.
• Full reproduction — run the pipeline in order from the raw sources (nb0 → … → nb13). This downloads hundreds of GB and takes substantial compute; run it in stages.
• Partial — any stage runs once its inputs exist (e.g. the classification notebooks once the event sets from stage 11 are present).

Raw RIPE RIS RIBs (~3.4 TB across the study) and case-study updates (~144 GB) are not redistributed; they are publicly available from RIPE RIS and fetched by the scripts.

Citation

@inproceedings{martinez2026circuitbreaker,
  author    = {Mart\'inez-Durive, Orlando E. and Chariton, Antonis and Fiore, Marco},
  title     = {The Circuit Breaker That Cried Wolf: Measuring BGP Maximum-Prefix Exceedance and Connectivity Impact},
  booktitle = {Proceedings of the 2026 ACM Internet Measurement Conference (IMC '26)},
  year      = {2026},
  location  = {Karlsruhe, Germany},
  publisher = {Association for Computing Machinery},
  address   = {New York, NY, USA},
  doi       = {10.1145/3777912.3839785}
}

License

Released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.