When a network path breaks
Trace how a missing link can change a route, why recovery takes time, and what the small interactive scenario leaves out.
Start with a working connection
Imagine a device reaching a service through an access network and several routers. The scenario shows two possible paths between shared endpoints. Its nodes and links are invented for teaching, not copied from a provider’s network. A drawn alternate path represents an assumed usable connection. The drawing alone does not establish independence of physical infrastructure.
The topology contains two assumed paths between shared endpoints.
An alternative path creates a possible recovery option.
Read the complete journey
- 1
Start with a working connection
Imagine a device reaching a service through an access network and several routers. The scenario shows two possible paths between shared endpoints. Its nodes and links are invented for teaching, not copied from a provider’s network. A drawn alternate path represents an assumed usable connection. The drawing alone does not establish independence of physical infrastructure.
Notice: The topology contains two assumed paths between shared endpoints.
Explore The access network
Why: An alternative path creates a possible recovery option. - 2
Remove one link
A link on the preferred route becomes unavailable. Packets cannot simply jump across the missing connection. Equipment must detect the failure or learn that the route is no longer usable. Some packets may be dropped or delayed while network state changes; our static switch does not model that time. Detection is itself a process.
Notice: Packets cannot cross the unavailable connection.
Explore Routers & forwarding
Why: Forwarding needs an operational link and current reachability state. - 3
Re-evaluate reachability
Within and between networks, routing mechanisms update reachability information according to their protocols and policies. An alternate physical connection only helps if an acceptable route exists over it. Business arrangements, configuration, filtering, and capacity can all affect whether the alternative can actually carry the traffic. Physical availability and acceptable reachability are distinct conditions.
Notice: New route information remains subject to policy.
Explore Network interconnection
Why: A physically connected alternative must also be permitted and usable. - 4
Use the surviving path
In this illustration, one alternate route remains available, so forwarding follows it after the assumed update. The route contains a different sequence of links. The model reports connectivity and route choice only. It does not infer delay, bandwidth, reliability, or geographic distance from how the drawing looks. Its capacity is a separate question.
Notice: The surviving route uses a different set of links.
Explore Routers & forwarding
Why: Forwarding can continue when an acceptable connected path remains. - 5
Recover what the connection can recover
Reliable transport can detect missing data and retransmit it if communication continues. Recovery costs time and can fail if the connection remains interrupted or exceeds protocol limits. Packet loss does not always require restarting a whole page, but an alternate path does not promise a seamless experience either. They need a viable connection.
Notice: Endpoints may retransmit data that was lost.
Explore TCP, QUIC & reliable delivery
Why: Recovery mechanisms can repair some delivery gaps after connectivity returns. - 6
Check the shared dependencies
The destination service still has to work. If both paths depend on the same failed access link, facility, router, power supply, or server, path diversity elsewhere may not help. Resilience is a property of the complete dependency chain, not simply the number of lines between two points on a diagram.
Notice: Both routes can still depend on the same endpoint or facility.
Explore Web servers & HTTP
Why: Resilience requires examining shared failures beyond the highlighted links.
Why do two drawn routes not prove that service will survive a failure?
Reveal the explanation
They may share physical infrastructure or endpoints, and the alternative still needs valid routing, permission, and enough capacity. Detection, convergence, and transport recovery can also interrupt service.
Sources & further reading
Original explanations informed by these sources. Diagrams are conceptual; local systems differ.
- RFC 1812: Requirements for IP Version 4 Routers IETF / RFC Editor; F. Baker (editor) · 1995-06 · Source & rights record
- RFC 4271: A Border Gateway Protocol 4 (BGP-4) IETF / RFC Editor; Y. Rekhter, T. Li, and S. Hares (editors) · 2006-01 · Source & rights record
- RFC 9293: Transmission Control Protocol (TCP) IETF / RFC Editor; W. Eddy (editor) · 2022-08 · Source & rights record
- The Internet Way of Networking: Defining the critical properties of the Internet Internet Society · 2020-09-09 · Source & rights record
- RFC 9000: QUIC — A UDP-Based Multiplexed and Secure Transport IETF / RFC Editor; J. Iyengar and M. Thomson (editors) · 2021-05 · Source & rights record
Another way through the system.
Load a page on your phone
A request across a network of networks.
Why the second visit can be faster
The trip that does not need repeating.
Send a photo while moving
Radio is the first part of a much longer trip.
Why a fast link can feel slow
Separate capacity from the waiting between actions.