How the Internet Still Works When Cables Break
The Internet's Invisible Redundancy
The modern internet survives physical damage (cable cuts, equipment failures, natural disasters) that would have destroyed it in the 1990s. This resilience isn't accidental—it's engineered through intentional redundancy, sophisticated routing protocols, and geographic distribution.
How Cables Are Protected and What Happens When They Break
Submarine Cable Networks: The internet is primarily physical cables, especially transoceanic connections. Approximately 500 submarine cables span global oceans, transmitting 99% of international data. These cables are surprisingly vulnerable: ship anchors, fishing equipment, earthquakes, and volcanic eruptions regularly damage them.
Example: In January 2022, the Tonga volcanic eruption severed the Tonga Cable System, cutting Internet to entire islands for weeks. In November 2022, the SEA-ME-WE-5 cable was cut near Egypt, causing disruption across Asia and Africa.
Resilience Through Diversity: Instead of single cables between regions, companies now deploy multiple cables on different routes. Analysts recommend using "at least four diverse cables" on major routes (APAC) to achieve "five nines" (99.999%) uptime.
Bangladesh's 2022 experience exemplified this: the country invested in regional interconnection, content distribution, and cable redundancy. When major submarine cables were damaged, alternative routes maintained connectivity while other countries with single-cable connections went offline.
AI-Powered Monitoring: Modern systems use fiber-optic sensors and AI algorithms to monitor cable health in real-time, enabling prediction and prevention of failures before they occur.
The Border Gateway Protocol: Internet's Immune System
How BGP Works: The Border Gateway Protocol (BGP) is a routing protocol—a "decision-making system" determining how data packets traverse the internet. Thousands of independent network providers (ISPs, cloud providers, corporations) use BGP to announce "I can reach destination X via my network" and choose optimal paths.
Automatic Failover: When a network link fails, BGP detects the failure in seconds and automatically withdraws the route announcement, forcing traffic to alternative paths. This happens without human intervention—routers continuously monitor path health and adjust routing automatically.
Example configuration: If your primary data center link fails, BGP health checks (IP SLA tracking) detect the failure, trigger a route withdrawal, and traffic automatically reroutes through secondary links within seconds.
The Real Latency Issue: While failover is automatic, it's not always optimal. Traffic may reroute through slower, congested paths, increasing latency temporarily. For financial trading (where microseconds matter), this is unacceptable; for web browsing, 50-100ms latency increases are transparent.
Network Redundancy Mechanisms
Multi-Provider Strategy: Large organizations don't rely on single ISPs but maintain connections to multiple providers. If one provider's network fails, traffic automatically reroutes through others.
Geographic Distribution: Rather than centralized infrastructure, systems distribute across regions. Data centers in Singapore, Sydney, London, and São Paulo can serve users worldwide, with traffic automatically routing to nearest location.
If one region goes offline, traffic simply doesn't route there—users in other regions are unaffected.
Redundant Hardware: Dual modem routers maintain connections to two different carriers. When primary connection fails, secondary automatically activates. This principle extends globally through data center redundancy, power supplies, and network interfaces.
Open Shortest Path First (OSPF) and Virtual Router Redundancy
OSPF for Internal Networks: While BGP routes between independent networks, OSPF routes within organizations. OSPF dynamically calculates optimal paths and reroutes traffic on failures.
Virtual Router Redundancy Protocol (VRRP): VRRP enables multiple physical routers to act as single virtual router. If primary router fails, backup automatically assumes its identity without service interruption. This is "active-passive failover"—backup standing ready, taking over instantly if primary fails.
Real-World Case Study: Resilience Hierarchy
Country-Level Resilience: Countries with multiple international cable landing sites and numerous providers experience resilience during submarine cable cuts. Those with single cable landing sites go completely offline.
West Africa's 2022 cable damage exemplifies this: countries with multiple transit providers experienced minor slowdowns; those with single provider went offline.
Provider-Level Resilience: Major cloud providers (AWS, Azure, Google Cloud) engineer availability zones—geographically separate data centers with separate power, network, and cooling. Entire region failures don't affect services in other regions.
Network-Level Resilience: BGP routing ensures that even if direct path is unavailable, alternative paths are automatically explored. The internet has no single point of failure at network level.
What Breaks This Resilience
BGP Hijacking and Configuration Errors: Most internet outages don't result from cable cuts but from BGP misconfigurations or attacks. An ISP accidentally announcing routes for major companies' IP addresses can redirect traffic—intentionally or accidentally hijacking their services.
Correlated Failures: When multiple cables in same location are damaged (earthquake, ship accident), redundancy fails. Japan's 2011 tsunami damaged multiple submarine cables simultaneously, causing outages despite supposed redundancy.
Power Grid Failures: Sophisticated routing can't help if the data center loses power. This explains why data centers over-invest in power redundancy—it's the single point of failure most likely to be hit.
Emerging Vulnerabilities
Satellite Internet Fragility: As internet increasingly depends on satellite (Starlink, Amazon Kuiper), new vulnerabilities emerge. Satellites can be disabled by space debris, attacks, or solar events. Unlike submarine cables (thousands deployed redundantly), satellites are expensive enough that operators can't deploy true redundancy.
AI Vulnerabilities: AI-based network monitoring enables prediction of failures, but also creates new attack surfaces. If AI systems can be fooled into withdrawing healthy routes or advertising false routes, they become attack vectors.
Common Myths
Myth 1: "Internet is decentralized and has no single point of failure"
Reality: While the backbone has redundancy, specific chokepoints exist. TSMC producing 90% of advanced semiconductors, NVIDIA controlling 95% of AI chips, and 3 companies controlling most global internet traffic all represent critical dependencies.
Myth 2: "The internet automatically fixes all failures"
Reality: Automatic failover is impressive but imperfect. Failover takes seconds to minutes; alternative paths may be congested; coordinated attacks can overwhelm redundancy.
Myth 3: "Internet routing is simple"
Reality: Global routing is extraordinarily complex. BGP operates on billions of route announcements from thousands of independent providers. Misconfiguration in one provider can cascade globally.
Why It's Trending Now
Internet infrastructure failures are becoming more visible as digital services become essential. The 2022 submarine cable damage events highlighted infrastructure fragility. Geopolitical tensions (Taiwan's strategic importance for chip manufacturing and cable landing sites) make infrastructure resilience a national security concern.
Future-Proofing
Regional Resilience: Countries building diverse domestic hosting, multiple international cable landing sites, and local ISP competition experience better resilience.
Fiber Expansion: Submarine cable capacity is being aggressively expanded. New cable projects add redundancy and capacity, reducing bottlenecks.
Open-Source Monitoring: Communities developing open-source network monitoring tools are improving visibility into infrastructure health.
Conclusion
The internet's resilience isn't magical—it's engineered through intentional redundancy, sophisticated routing protocols, and geographic distribution. However, resilience has limits: correlated failures, power outages, and concentrated dependencies (TSMC, NVIDIA, specific cable routes) remain vulnerability points. As internet infrastructure becomes essential infrastructure, investment in resilience and redundancy becomes increasingly critical.