4 ms·

Why is it not more trivial to auto-route around those kinds of events?

• walrus01 8y ago
  In many cases BGP does its job just fine at layer 3. But budgetary constraints and financial reality for the cost of really huge interconnection mean that a cut may result in a 100 Gbps PNI peering session between two ISPs filling up and flat-topping its traffic chart (very, very bad) when the traffic re-routes. One example is that as a result of today's events there was apparently major packet loss and congestion between HE and Comcast in the NYC area. Or when traffic re-routes, a major ISP will temporarily congest its own intra-AS city-to-city links on a few key paths, for their own router-to-router interfaces.
  • drewr 8y ago
    Ah, makes sense, you still need the ability to handle the rerouted traffic.
    • joering2 8y ago
      Well precisely. And not like Comcast doesn't have money aka "But budgetary constraints and financial reality...". The reality is that they do annual surveys and find out spots on map where cutting happens very rarely and in those places they won't have any redundancy running - something, you know - the whole internet is based off of. So no, two cuts like this are not something out of ordinary. The fact Comcast is cheap and don't have redundant pipes/contracts to operate secondary backup lines is simply being cheap and saving money on their part.