As all our customers know, earlier tonight the core1.iad1 router replacement (12008/PRP to 12410/PRP-2) and upgrade of certain ports of it to 10G was completed. However, what deserves to be noted is how smoothly the upgrade went. Even though, temporarily, DCA2 and DCA3 data centers were cut off from the rest of the network, and the SEA2 data center, SEA1, PAO1, SJC1, and LAX1 sites from DCA2 and DCA3, all communication between DCA2, DCA3 and the rest of our network continued uninterrupted. This is a feat that, to the best of our knowledge, no other IP backbone has been able to accomplish, even in most dire of circumstances. Namely, ensuring that network connectivity is uninterrupted even when local network connectivity between some backbone sites is lost, having complete redundancy of not just transport but also transit – and with the advanced and failure-safe configuration in place.
This was the first real world performance of our 9 years ago designed unique system of numerous checks and balances and advanced, nearly never used by other networks, configuration directives, designed to automatically make one network site that is cut off from the rest communicate to the rest of our network over the public Internet – that is, accept our own AS routes in our own AS, but, only so when they are not reachable via the hopone.net backbone itself.
While this advanced configuration – that no other network has been able to architect and implement safely and transparently – was in place for nearly a decade, we had never had any network outage or issues that would have invoked it. Until, that is, this night’s scheduled replacement (upgrade) of the core1.iad1 router. And, much to our delight, it worked flawlessly just as designed. And, unlike any other backbone that would have had massive outages and customer impact in such a case, there was just a few second blip as various IP networks recalculated their routes to us, and all continued running normally.
This is a truly amazing achievement (and I am never one to use such words lightly) worth noting, showing how the hopone.net backbone is truly far ahead of the rest and is more fault tolerant by its numerous redundancy and failover layers than any other backbone.