The latest story is summarized as a government contractor was installing a new router as part of a network upgrade, and the router generated excessive amounts of traffic that existing points in the network couldn't handle
HPE Juniper says the routers were not malfunctioning and did what they were configured to do
The implication is that the router was misconfigured in some way, but no one really knows. Given how government contractors work I buy that story, that they were doing production changes in the network at 3PM on a Friday and had a bad config
No indication at all that it was that new AI tool
It would be really rare or impossible for a router to generate enough IP traffic on it's own (eg sourced from the router itself) to saturate either the fiber line, or the general IP processing ability of other routers.
I feel like it has to be an issue with whatever routing protocol they are using. So, low traffic in absolute (bps) terms, but something expensive for other routers to process.
I hope they provide a more detailed explanation.
I doubt that the router itself generated that traffic, I suspect either there was some kind of loop created that amplified broadcast traffic, or too many VLANs on a certain link and it forwarded more traffic than was intended, etc. There are a lot of scenarios where after a game of telephone makes it from the engineer on the ground to the Secretary of Transportation that would be close enough for him to explain it the way it was
Re:
by meeshmuesh on Tuesday November 17, 2026 @04:56PM
(#50538796)
It strains credulity that their network is designed such that a poorly inserted cable or misconfigured vlan would being the network down. If the network was that badly designed there would be constant issues.
Re:
by lallysingh on Tuesday November 17, 2026 @05:54PM
(#50539417)
A well designed network can get in bad shape when a router is poorly configured. And we don't know how well or poorly designed the network really is.
Well, given what happened last week it's very possible the network is very badly designed
You'd be surprised how bad things can be and still skate along without breaking for a long long time
Re:
by cheesecakegood on Tuesday November 17, 2026 @09:17PM
(#50541039)
Even more bizarrely, today it was reported that the FAA literally had someone supervising the installation, checked the work, and signed off on it, and the initial tests were perfect. So something isn’t adding up. It could be someone is pants on fire lying to save their ass, but it could also be something complicated not yet identified.
Re:
by ShinyLeftPad on Wednesday November 18, 2026 @01:56AM
(#50542768)
> No indication at all that it was that new AI tool
Except the timing exactly coincides...
I have been explaining the FAA outage to my less technical friends this way: Think of the incident like a hyper, fast-talking New Yorker -- who just drank two pots of coffee -- walking into the quiet back room of a Southern saloon and started asking 250 questions a second at a few slow-talking good ole boys used to passing one polite sentence at a time down a single-lane dirt road. That dog just ain't gonna hunt.
Was the issue reliably identified then? Last I remember they swore it was definitely not the new SMART thing and failed to provide a satisfactory cause.