[usernam33]

Imagine it being 8pm, you sit in a operating room of germanys largest internet knot with up to 6,027 Tbit/s throughput and then the power goes down.

I am concerned that this even happened in a place important like this.

pic related: http://img.pr0gramm.com/2018/04/09/9286baa4b2261184.jpg

The article states internet was down for users at around 10pm (I expirenced it live). So they might have been on generators for two hours (speculation).

Shouldn't we decentralize more?

[virtuallynathan]

Only those peers who are single-homed to Interxion FRA5 / DE-CIX6 will have experienced an outage. The whole of DE-CIX isn't down.

[AdamJacobMuller]

> Shouldn't we decentralize more?

> The whole of DE-CIX isn't down.

Largely agreed with you, though as general advice I would say that redundancy by adding more exchanges to your network serves you better than adding diverse connections to a single exchange.

[virtuallynathan]

Yep, that makes a lot of sense, especially since 10G/100G waves and such on the FLAP route are so inexpensive. Getting a remote peering port on AMS-IX or something is probably a good move.

[mmarx]

> The article states internet was down for users at around 10pm (I expirenced it live). So they might have been on generators for two hours (speculation).

More likely, the timestamp in the picture is UTC, which corresponds to 22:00 CEST, cf. [0]

[0] http://status.core-backbone.com/index.php?id=1972

[walrus01]

Well, now we know that significant portion of that facility are 1+0 for power systems. In a true 1+1 setup with air gap between A and B side power systems this wouldn't have happened.