• A 14-minute Cox outage centred on Chicago affected partners and customers across multiple countries
  • An eight-minute Verizon Business outage two days later also reached customers across several regions

The fact

Cisco ThousandEyes recorded two short US network outages with effects far beyond the nodes where the problems were observed. On 9 September, Cox Communications experienced a 14-minute outage that appeared centred on Chicago. ThousandEyes reported affected partners and customers in the US and 12 other countries across North America, South America, Europe, Asia, Africa and Australia.

On 11 September, Verizon Business experienced an outage lasting about eight minutes. It initially appeared centred on nodes in Chicago, Seattle, Los Angeles and San Jose, while affected customers and partners were observed in the US and 16 other countries.

The monitoring report did not identify the root cause of either event or publish the routing and commercial relationships connecting every affected network.

The assessment

The short duration matters less than the number of downstream networks that depended on the affected paths. Internet traffic can cross several providers between an application and its user, so a problem inside one operator can become visible to customers that are geographically far from the affected nodes.

That does not mean every affected customer bought service directly from Cox or Verizon, and the ThousandEyes data does not reveal every underlying contract or route. It does show that geographic distance does not provide protection when networks share upstream or interconnection dependencies.

For BTW readers, resilience therefore depends on where supposedly separate paths converge. Two connections can enter a building through different cables and still become dependent on the same upstream network further along the route.

What to watch

Watch for any customer-facing incident reports or provider disclosures explaining the causes of the Cox and Verizon events. Repeated outages involving the same cities or upstream paths would make the dependency easier to identify. Routing and active-path measurements can also show whether supposedly diverse enterprise connections repeatedly converge on the same networks.