A backup internet route can stay online and still slow you down

Photo: Stefan Funke, Switch Rack, 2011, via Wikimedia Commons / Flickr. CC BY-SA 2.0. Original photograph unaltered; creator retains copyright. Archival 2011 photograph of DE-CIX-branded switch racks. It shows real exchange infrastructure, not the 2026 research experiment, and establishes no current route, installation or performance. The original photograph is unaltered. [Photo source](https://commons.wikimedia.org/wiki/File:DE-CIXGERMANY-SwitchRack_(6218137120).jpg). CC BY-SA 2.0.

A website can remain reachable after an internet route disappears and still become less pleasant to use. That distinction matters when a small business buys connectivity, a publisher serves readers in several countries, or a team depends on a cloud application. A promise of a backup connection invites a second question: what experience does that alternative actually preserve?

A research account published on Internet Society Pulse on 3 November offers a reason to ask. Weitong Li reports that withdrawing an individual internet exchange point path from a highly redundant anycast network did not cause persistent loss of connectivity, but could leave affected networks with worse latency. The account describes suppressing BGP communities at global anycast content-delivery-network vantage points. It reports that exchange size was only weakly related to the impact. Read the author's account.

That is a reported experiment with a particular kind of network, not a finding that every subscriber has a safe backup. PING has not repeated the measurement or inspected a full underlying paper or dataset. The post does not establish the performance of a reader's broadband line. Treat its conclusion as a reason to examine alternatives, rather than a ranking of providers or exchanges.

The exchange is a meeting point

An internet exchange point, or IXP, is infrastructure where separate networks connect and exchange traffic. Internet service providers, mobile operators and content delivery networks can all participate. The Internet Society's June 2020 explainer describes the physical requirements, including switching equipment, a suitable location, power, cooling and people to run it. It also emphasises agreements and trust between participants. A rack is only the visible part of that cooperation. IXP explainer.

Routing is another part. BGP is the protocol through which separate networks exchange information about which destinations they can reach. Its base specification describes reachability information and the network paths that information traverses, allowing policy decisions as well as loop prevention. Knowing that a destination remains reachable does not itself describe how quickly an application will respond. The January 2006 specification is background here, with subsequent updates listed by the RFC Editor. It is not a new release or configuration recipe. BGP specification.

The operator whose equipment appears in the archival photograph gives a concrete example of this interconnection layer. DE-CIX's current Frankfurt page describes route servers that facilitate the exchange of BGP announcements between peers. That is the operator's service description. PING has not tested it, and the 2011 image cannot confirm today's equipment or the route used by a particular customer. DE-CIX Frankfurt.

Start with the places your service must reach

For a London publisher with readers in Lagos or Nairobi, the practical question is about those readers reaching the actual service. A city name on a sales page is an incomplete answer. Equally, an African business serving local customers should not treat a measurement made elsewhere as proof of its own experience. Those are examples of how to frame a check, not a claim that any named city or operator was measured by PING.

Ask the provider which service the resilience promise covers and what happens when the usual path is unavailable. Ask what evidence it can supply about the alternative, for which destinations and under what conditions. Keep the answer alongside the contract and support process. If an application is important to the business, agree how its behaviour will be checked during a planned test. A technically qualified operator should manage any routing change; this article is not an instruction to withdraw live routes.

Separate the outcome you need from the infrastructure label. Staying connected, maintaining an acceptable application response and reaching support are different questions to document. A useful review therefore starts with the work people need to finish, then asks the provider to connect its technical explanation to that requirement. The research gives that conversation a sharper starting point without promising a particular speed or a universal solution.

The sensible payoff is a better description of the backup you already pay for. Record its scope and the evidence behind it before making it part of a continuity plan. For the related question of whether a network footprint becomes a usable connection, read PING's coverage and connection explainer.