I discovered caching CDNs were throttling my everyday browsing

Caching CDNs and ISP routing quirks can make large parts of the modern web feel “broken,” even when basic connectivity and speed tests look fine. Commenters argue the root cause is usually overloaded or poorly managed peering between residential ISPs and major CDNs, sometimes compounded by traffic shaping, MTU issues, or “optimizer” middleboxes, rather than CDNs deliberately throttling users. The thread broadens into a comparison of UK ISPs’ peering quality, support, and bandwidth caps, alongside the practical challenges of self‑hosting sites that suddenly receive heavy traffic.

Diagnosis and Root Cause Theories

  • Many commenters think the issue is not intentional CDN throttling but overloaded or mis‑provisioned links between the ISP and major CDNs (especially Akamai).
  • Others propose classic ISP traffic shaping or “optimizer” middleboxes that treat long TCP flows (“elephant flows”) harshly, but don’t affect VPN (UDP/WireGuard) traffic the same way.
  • Some note that VPNs or IPv6 changing the path and/or CDN POP strongly suggests routing/peering issues rather than end-site problems.
  • A minority suggest local issues like MTU/MSS or PPPoE misconfiguration could cause partial breakage, though others argue the packet captures don’t clearly show fragmentation problems.

ISP Quality, Zen vs Alternative Providers

  • Several users report similar performance issues on the same ISP, including slowdowns to CDNs and certain sites.
  • Others say their service from the same ISP is solid, including high‑speed FTTP and IPv6, implying the problem may be regional or related to specific routing changes.
  • A competing UK ISP is repeatedly cited as having strong technical support, detailed line monitoring, and an ability to pressure Openreach effectively.
  • Some argue support quality matters because Openreach infrastructure is often unreliable; others think line stability is more important until something breaks.

Peering, CDNs, and Net Neutrality

  • Multiple network operators in the thread emphasize that performance often hinges on private peering capacity and where CDNs and ISPs choose to interconnect.
  • They argue CDNs are incentivized to deliver bits as fast as possible; problems usually stem from ISPs under‑investing in peering or backhaul.
  • There is brief discussion of net neutrality: concern about ISPs and “content taxes,” but also claims that the real fairness issue is poor ISP investment, not CDN “prioritization.”

Traffic Shaping, “Unlimited” Plans, and Caps

  • Several comments describe long‑standing ISP practices: shaping long downloads, burst‑then‑throttle behavior (especially on cable/DOCSIS), and using congestion signals to force lower video bitrates.
  • Debate over “unlimited” vs capped plans: some say terabyte‑scale caps are fine for most users; others call 1 TB/month inadequate in 2023.
  • Some note that so‑called “unlimited” plans often hide fair‑use clauses or traffic management, while others give regional examples where high‑volume usage proceeds without consequence.

IPv6, MTU/PPPoE, and Low‑Level Tweaks

  • A number of technical suggestions appear: try IPv6, test with hotspot, lower MTU or clamp MSS, or ensure PPPoE overhead is handled correctly on DSL.
  • Tools like browser‑based PMTU tests and speed sites tied to video CDNs (fast.com, speed.cloudflare.com) are suggested to detect shaping more realistically.

Home Hosting, CDNs, and HN Traffic

  • The blog was initially self‑hosted on a low‑upload home line and was quickly overwhelmed (“hug of death”).
  • Commenters debate home hosting viability: some say single‑digit Mbps is fine for personal sites; others argue it degrades quickly with images or sudden spikes.
  • The author eventually fronted the site with a CDN, with some irony given the article’s topic, and was advised to tune CDN caching rules and optimize images.