New FCC Broadband Standards Should Consider Latency
Calls to update U.S. broadband standards are shifting from a narrow focus on download speeds to include latency, jitter, and reliability, which more directly affect video calls, gaming, and general responsiveness. Commenters argue that regulators should define clear, testable latency metrics—ideally under real-world load and to well-connected exchange points or measurement servers—while warning that averages and idle measurements can hide problems like bufferbloat and micro‑outages. Many also note that transparent “nutrition labels” for ISPs will only go so far without stronger competition and incentives for better peering, last‑mile engineering, and infrastructure in rural areas.
Scope of “latency”: to where?
- Strong disagreement on the proper reference point: ISP first hop, ISP edge, nearest IXP, Tier‑1 backbone, CDNs, or specific public sites.
- Many favor measurement to a well‑connected IXP or standardized FCC test servers, not arbitrary on‑net endpoints that ISPs can game.
- Concern that latency is inherently end‑to‑end and partly outside ISP control, but others argue ISPs can improve it via better peering and routing, so they should be partially accountable.
Working latency vs bandwidth
- Repeated claim: today’s apps are typically not bandwidth‑limited; they are limited by latency and jitter, especially under load.
- “Working latency” (latency during active upload/download) is seen as the meaningful metric; idle latency alone is called misleading.
- Examples show networks with low idle latency but multi‑second latency under load due to bufferbloat.
How to measure latency and reliability
- Suggestions:
- Latency to FCC “Measuring Broadband America” servers.
- Latency to IXPs, major CDNs, or a set of .gov sites.
- Metrics like mean/median, high percentiles (p95/p99), standard deviation, and tail indicators (“glitches per minute”).
- Include packet loss and short “brownouts,” not just uptime percentage.
- Debate over measuring latency under full load vs best‑case idle conditions; some argue full‑load reveals broken buffering, others say it confuses users.
Real‑world experiences and technologies
- Reported latencies: low single‑digit ms on good fiber, tens of ms on cable/DSL/4G/5G, 20–50 ms on Starlink, 600–1000 ms on geostationary satellites.
- Users highlight jitter and micro‑outages as bigger problems than peak Mbps, especially for SSH, VoIP, and video calls.
Regulation, labels, and market structure
- Some welcome FCC “nutrition labels” including latency, availability, and variability metrics.
- Others worry detailed requirements add legal/administrative burden that favors incumbents and doesn’t fix lack of competition or restrictive building leases.
Web performance vs network performance
- Several note many “slow” sites are dominated by page bloat, trackers, and dependency chains, not raw network latency.
- Disagreement on whether this is primarily a web‑development problem or a network/ISP problem.