Some interesting stuff I found on IX LANs#

Ben Cartwright-Cox (bgp.tools)

• Why: Every IX fabric is a shared Ethernet broadcast domain, and what actually appears on it versus what the AUP forbids is a recurring L2 security and hygiene problem. This is empirical, at scale — 100 peering exchanges observed from one vantage point — and directly relevant to anyone running or consuming IX infrastructure.
• What: A survey of "bad" broadcast/multicast traffic bgp.tools has observed across ~100 peering exchanges' LANs, an assessment of the real-world impact of that traffic, and concrete suggestions for how exchanges could police it (from abstract only; the program page notes the slides are an evolving version previously shown at EuroIX).
• How: bgp.tools passively observes IX LANs as part of its normal data collection, so the findings are longitudinal, real traffic rather than a lab reproduction.
• Where applicable: Broadly applicable — the failure modes of shared L2 domains (stray ARP/ND, unwanted multicast, leaking protocols) are not IX-specific; the same lessons apply to any large shared switching fabric. The mitigation proposals are aimed at IX operators, but members can use the findings to check their own edge hygiene.