What Happens When an Undersea Cable Is Cut: BGP Explained

When a submarine cable is cut, the internet does not go down. It reroutes. That process, driven by a 30-year-old protocol called BGP, takes anywhere from a few seconds to several minutes — and during it, packets are dropped, latency spikes, and some connections fail entirely.

What Is BGP?

BGP (Border Gateway Protocol) is the routing system that tells internet traffic how to get from one network to another. Every major internet provider, cloud company, and large enterprise runs BGP to advertise which IP addresses they own and the best paths to reach them. There is no central router for the internet. BGP is how thousands of independent networks agree on routes to each other in real time.

What Happens the Moment a Cable Is Cut

When a fiber strand in a submarine cable breaks, the routers at both ends stop receiving data. Within milliseconds, a protocol called BFD (Bidirectional Forwarding Detection) picks up the link failure. With BFD configured correctly, failure detection completes in under one second.

Without BFD, the default BGP hold timer is 180 seconds. That means three full minutes pass before BGP even registers the cable is gone. Most major internet providers have deployed BFD, but misconfigured or older networks still use default timers.

Once BGP detects the failure, two things happen simultaneously:

• Route prefixes reachable through the broken cable are withdrawn from the routing table
• BGP searches for alternate paths to the same destinations across remaining cables

This process is called BGP convergence. During convergence, routers exchange update messages, calculate new best paths, and gradually stabilize. Depending on network size, configuration, and how many prefixes are affected, full convergence takes 30 seconds to several minutes.

What Developers See During Convergence

During convergence, applications experience:

• Packet loss: in-flight packets routed through the failed link are dropped
• Latency spikes: traffic reroutes over longer geographic paths, adding 50-200ms or more
• Connection resets: long-lived TCP connections drop if the route changes mid-session
• Route flapping: routes appear, disappear, and reappear as BGP stabilizes

For a web request or short API call, this usually means a single timeout that a retry resolves. For database replication streams, video conferencing at scale, or financial trading systems, it means visible, sometimes costly disruption.

How Traffic Reroutes in Practice

When Red Sea cables are cut, traffic between Europe and Asia cannot take the direct route through the Middle East. BGP automatically shifts it to alternate paths:

• Around Africa via Cape of Good Hope route: adds 150-200ms round-trip latency, significant congestion on cables not designed to carry the full load
• Trans-Pacific route: adds roughly 300ms round-trip for Europe-Asia traffic
• Satellite capacity: last resort, very limited bandwidth relative to cable volumes

The 2022 and 2024 Red Sea cable damage events showed this sequence in action. Internet in parts of South Asia and East Africa slowed noticeably for weeks because congestion on alternate routes exceeded their available spare capacity.

What Developers Should Actually Configure

Most applications are not built with regional cable failure in mind. These changes reduce the blast radius significantly:

• Increase TCP keepalive settings to detect dead connections faster (the 60-second default is too slow during cable events)
• Set aggressive timeout values on API calls — a 30-second timeout during convergence will queue connections rather than fail fast and retry with exponential backoff
• Implement retry logic with exponential backoff for all external calls
• Monitor latency per region continuously with Pingdom, Datadog Synthetics, or Catchpoint — a sudden 200ms increase to a specific region is the first signal of a routing event
• Do not rely on a single CDN point of presence for critical traffic — distribute across CDNs with different upstream providers

For architecture guidance on surviving a full regional outage, see Your Cloud Region Just Went Offline: What Happens Next.

Key Takeaways

• BGP — Border Gateway Protocol, the routing system that reroutes internet traffic after cable failures across thousands of independent networks
• Under 1 second — BFD detection time for link failures when properly configured; 180 seconds with default timers
• 30 seconds to several minutes — typical BGP convergence time after a major cable cut
• 150-200ms — added latency when Red Sea traffic reroutes around Africa via the Cape of Good Hope
• For developers: increase TCP keepalive timers, implement retry with exponential backoff, monitor latency per region, and set tight API timeout values — these changes matter most during cable convergence events
• What to watch: EIG, IMEWE, and SEA-ME-WE 5 cable systems are the critical paths for Europe-Asia traffic through the Middle East. Any simultaneous damage to multiple systems forces rerouting through already-congested alternates