This is a pretty classic network operator story: go to great lengths to provide for physically diverse paths, then not notice when your provider refactors and grooms them onto the same bundle.
Over ten years ago, my employer was spinning up a new DC across the state and had three links between it and the primary DC. Two were pretty direct, but we needed a third because at one point in the 300 mile path, the two main links went within 400 meters of each other.
So how national-security-adjacent critical systems like an airport system doesn't have a larger set of backup links, and validates that they are geographically separate up until the connections, and have realtime alerting on the connection status, is surprising to me.
> validates that they are geographically separate up until the connections
This isn't relevant in this case. The problems occurred at different places. The backup fiber was, separately from the issues with the primary system, cut by a construction crew. It's not two fiber lines both cut at the same spot.
Similar events happened in Europe disguised as thieves stealing fiber optic. This does not have any sense economically, as the value in the market is zero so... either is an honest accident and is cleared in a few days, or is sabotage
>disguised as thieves stealing fiber optic. This does not have any sense economically, as the value in the market is zero
Do you honestly think that crackheads think that far in advance?
I've seen fiberoptic cables stolen from 2 (city) jobsites in the last 5 years, once by tweakers later caught trying to sell them as scrap copper and the second thief was never caught.
This happened even though the spools had big signs on them saying "Fiber Optic Cable - NO COPPER".
This used to happen more often in the past, when copper thieves mistook fiber for conductor. There were cases where critical systems were shut down due to fiber optic theft, and the thieves were caught burning the sheathing to expose the... glass.
That's very uncharacteristic, especially given all those contingency requirements (backup policies, failovers, testing schedules etc.) imposed on corporations/companies in the wake of 9/11.
You actually have to ensure that someone isn't just checking boxes the tests passed and the tests actually passed.
I worked for a regional ISP that had a major outage when the redundant fiber provided by the telephone company was cut in one place triggering a full loss of connectivity. It also caused a massive 911 outage for 200,000 people as it isolated the 911 center from the city core.
Turns out the phone company didn't connect one side of the ring topology even though they certified they did. Needless to say lawsuits abounded.
And auditing that is expensive and you always get given hell from the assholes going “why am I paying for this if nothing has gone wrong”
The amount of shit I was given everytime I went through a checklist working at a fedramp certified company working with emergency alerts on the phone system made me prematurely gray.
I could feel the barely contained seething rage everytime I told the execs that the reason this release will take 3 days and not be instantaneous like your friends releases at a faang and that information made you embarrassed at your dinner party, is because you agreed to this process contractually years ago and now it’s a crime if I just sign off on it being ok without actually checking that it’s ok.
Yea, fangers are def not used to working on life critical systems. I personally have avoided it as much as possible, but not completely, for all the reasons you listed above. No need for me to accept that much responsibility for other peoples lives when I've been able to earn a living just fine not doing so.
There are plenty of people who are happy to check off "Backups" boxes, or talk about their backup strategy, or whatever - while hoping the day never comes when they actually need that tricky "Restore" feature.
You don't always have a copy of your hardware to restore onto. And the test's entire purpose is that you're not yet sure whether your restore will truly work. So you can't just run a backup and restore on your true prod system, because you're not sure it won't wreck it. So you need extra money to have a second system onto which you try to restore. If you don't have a lot of money, you will want to actually use your disks for storage, not to put them into a second testing server. Of course I'm not talking about very professional companies with super critical data. Just simpler smaller scale places or consumers.
No, even professional companies with critical data balk at this.
I worked at a company worth a few billion and the leadership balked when they told engineering they wanted a near instantaneous failover system and our department informed them that would require paying for a second environment that could be rolled over to.
It is rare to find leaders who can accept the cost of redundant infrastructure that is there for emergency backup.
What puzzles me is why they can’t accept it when they are perfectly fine with insurance costs and I can’t see much of a difference between the two when looking at a spreadsheet of costs other than possibly tax differences between the type of expenditure.
I've worked on backup/failure systems since the mid-90s, and I've found there's one universal truth: If you don't fully test your backup/failure system, you don't have a backup/failure system.
There's generally two wrong responses: (1) We spent a lot of money on 'blah blah blah', a lot of other companies use it, so yeah, we've got a backup/failure system. And, (2) inadequate testing - either, we tested 1 of 50 services, and it worked, so the whole system can be restored; or, we gracefully tested, and it worked, so it will obviously work during not-graceful incidents.
And the root cause of this is generally that no one gets promoted for implementing an adequate backup/failure system, or it's extremely rare.
Besides what the others have said, speaking from a neteng perspective, sometimes backup lines (and the core infrastructure in general) are engineered in such a way that it can't easily be tested properly without taking other things down, or manually rolling a truck specifically to test it in isolation with extra equipment.
Not saying that's what is going on here, just that it's possible.
Finding out your backup fiber has been cut, aparently some time
ago, only when you try to use it as the backup isn’t an issue of “aging infrastructure.” It’s an issue of letting completely incompetent people run your tech.
Building multiple diverse paths and monitoring for fiber cuts is not hard. Having 2 fiber paths is insufficient for even only moderately important workloads at a tech company. Overlapping fiber cuts happen. For something with significant economic and safety impact this is just crazy.
Sometimes the level of incompetence / lack of care in organizations like this astounds me. I understand issues like this can be complicated and systemic but it honestly makes me think very poorly of the technologists building these systems in government.
I've told this story so many times I've worn the corners off it. BT was giving a presentation to MSN WAN OPS about the new datacenter buildout in London for us. They kept going on and on about physical security, man traps, and that our fiber left the building on each side and didn't get close to each other for so many km away.
BT guy ends that part of the presentation with "the IRA will really have to get shit together to take you off the net"
This was during the troubles. Same planet, different world.
They had two diverse paths, but apparently did not have a (reliable?) process for ensuring the backup path was functional during fail-over. I imagine there will be some soul-searching over that.
Nothing I've ever seen or experienced with FAA indicates a lack of care; the parsimonious explanation is almost always that people who care a great deal are working within complex systems that don't always have a consistent or externally legible set of priorities. Or more intuitively, the failures we see are the "acceptable" ones versus the unacceptable ones (like planes falling out of the sky).
> Nothing I've ever seen or experienced with FAA indicates a lack of care
Hard to fathom someone still holding that view after the 737MAX disasters.
The NTSB has repeatedly and publicly criticized the FAA for failing to implement their recommendations, including as recently as last week (Amazon Prime crash resulting in 5 deaths). They've explicitly called out the FAA's failure to act as a contributing factor in crashes, multiple times.
The problem is with considering the second path a backup path. It is a common problem with things that are considered to be a backup. In contrast, both paths should ideally have been in regular use. As a general policy, for resiliency, one should always exercise all of one's routes/suppliers/vendors, even if some are suboptimal, not just the primary one.
I completely agree. My comment is not to suggest there wasn't a problem or failure (clearly there was one!), but that we shouldn't assume incompetence versus conflicting priorities under limited resources.
That sounds pretty plausible. But you still have to have enough internal competency to oversee a contractor, and ask the right questions/build proper requirements and verify they are meeting them.
Over the weekend, I got linked to https://admiralcloudberg.medium.com/reaping-the-whirlwind-in..., which is an exhaustive look at the causes of the first airspace collision in the US in several decades. After reading that, my conclusion is that, if the software being rolled out is what I think it is, it would be a borderline impeachable offense to not first trying it out in the DC airspace.
The fundamental problem with DCA, and one of the two main causes of the crash [1], is that they are required by political pressure to operate at a higher operational tempo than they can safely operate at. DCA has essentially 1½ usable runaways--large planes can only use the larger runway, and given the capacity restrictions, airlines have been pushing to use fewer small planes at the airport. This shift in plane size means the effective safe slot capacity has gone down, but people still keep citing the same number as justification for safe numbers, and the politicization of the issue has shut down everyone who complained that the actual traffic just couldn't be safely handled.
If DCA's theoretical slot capacity (36 landings and takeoffs each per hour) is to be reached one just one runway, you have about 100 seconds to go from plane 1 touchdown; exit the runway, letting plane 2 on to take off; plane 2's wheel leaving the runway, clearing plane 3 touchdown. That's doable, but with essentially 0 margin for error. Alternating the runways used for each landing would give you closer to 30s of margin, but if only 10-20% of the planes can use the alternate runway, you can't divert enough planes to use the alternate runway.
Now DCA and the FAA aren't stupid enough to actually schedule 36 landing slots every hour, but the problem is that because of a few various factors, what was scheduled as, say, 30 slots for an hour (giving 120 seconds between landings, probably sufficient margin) ends up being 12 slots used in the first half-hour and 18 slots used in the second half-hour, which means a lot of the actual operation ends up having no safety margin even though on paper you have sufficient margin.
One of the ways you can rectify that is to assign slots further in advance so that you don't get the bunching. Effectively saying "oh, if you leave right now, you'll arrive at 3:30 with three other planes, but if I hold you for 10 minutes, I can push you into a less busy arrival time." Doing this requires good, accurate prediction of the actual flight travel times, and my understanding is that this is what the new software is meant to provide.
[1] The other main cause is essentially that the DoD's aviation practices in the area is a giant clusterfuck that endangers lives, and unfortunately that sentence is not relegated to the past tense.
Honestly. Of all the places to deploy first I’m happy it’s that one.
The government politicians who approved this are frequent flyers out of there, so they’re at least getting their lines out on the line before the rest of the country if it proves ill managed or conceived.
I don't quite understand this sort of thing happening. Wasn't the whole point of the internet to be a self-healing network where we route around severed cables, etc?
Or is it that these ATC networks are their own air-gapped network with less redundancy? That just doesn't add up. Or maybe there was only one line going to the ATC, with no multiple "ISPs" like a datacenter would have?
* Yes, TRACON have their own dedicated links. You can look into ASTERIX and STARS to learn some of the cursed ways the data processing and dataflow work.
* There is supposed to be a primary and a secondary link, in this case the primary failed and the fail-over also failed. It's unclear from the reporting if they were damaged in the same incident or if the failover was not tested or monitored adequately.
The Philadelphia TRACON site has been notoriously unreliable and was supposedly improved in 2025, it's also unclear if these issue actually could stem from that implementation.
I’ve had enough double-redundant links fail - and I mean pretty thoroughly double-redundant, different provider, different media, different physical path - that I’m quite surprised something like the air traffic control system only has two links.
Contractually dual-path, or actually dual-path? My understanding is that there's enough infrastructure horse-trading going on behind the scenes that it's very difficulty to be certain that two circuits between points A and B don't share the same infrastructure somewhere in between.
First big oops of this form that I remember:
"In December 1986, the ARPANET had 7 dedicated trunk lines between NY and Boston, except that they all went through the same conduit -- which was accidentally cut by a backhoe. "
One thing that needs to be reviewed is if they are dual links or redundant links.
In any system where both sides of redundancy always carry traffic then loss of one link can cause congestion failure if any link fails. This is a very common means of failure in electrical networks that requires load shedding. Well, you can't load shed air traffic.
A more complex system that I like, but comes with it's own set of constraints and implementation issues is a system where both lines carry all the traffic at all times. This way the default state of the system is always working and your first failure isn't invisibly critical.
But this is very hard as we see in TCP when things get out of order and high latency creeps in. You have to manage a lot more state at the data level.
Not to be conspiratorial, but about an hour ago I saw this in The Economist: "Russia’s grey-zone attacks on Europe are growing more brazen"
[QUOTE:]
Russia has been staging covert operations and deniable provocations against European countries for many years. Since its full-scale invasion of Ukraine began in 2022 they have increased. These have included incursions by reconnaissance drones, hacking, arson, parcel bombs targeting cargo flights and plots to kill executives of leading European arms firms. In recent months the pace of such shadowy attacks appears to have risen again. The Russian operations have three main goals, according to analysts and security officials: trying to coerce European countries into halting their support for Ukraine; raising the costs of providing that support; and attempting to make NATO look powerless.
[/QUOTE]
And of course there've been "suspicious" cable cuts in the Baltic Sea and other European waters ....
As a backup for when ground communications are cut? As opposed to having no backup in that situation? Yeah, maybe we should, no matter who controls the company.
Fiber was all installed long after we knew locating it was important. I suspect it all had a tracer wire installed with it. Of course sometimes the tracer wire can break.
Before fiber a lot of things were put into the ground with no thought how you would find it again.
Indeed, tracer wire is usually bonded to a (sacrificial) magnesium anode, but it’s a crapshoot whether the tracer wire is still intact.
I sometimes hire directional boring and excavation contractors and if there’s any doubt as to where an electrical conduit or natural gas pipe is, I opt for the hydrovac truck to minimize risk.
Always bring a length of fiber optic cable when you're in the wilderness. If you get lost or stranded, bury the cable. Within a few hours, a backhoe will stop by to dig it up, and the crew will rescue you.
Local people know where they are because once upon a time a D8 drove through pulling a big orange spool of fiber conduit and passers by chatted about how another fiber was going in. Also around here the contractors seeded the disturbed ground with a wildflower mix (presumably to discourage weeds) so you can also look for straps of wildflowers.
And of course there are marker posts that say "buried fiber optic cable do not dig".
Once you get out farther from civilization these signs become less common and there are more situations where they can get removed.
Most cheap hired labor doesn't realize you shouldn't dig up the market posts and throw them in the dumpster on construction sites. Add a little bit of rain and suddenly the contractor showing up with an excavator that was told everything was properly marked, and can see markings in other places but not where they are digging, and disaster occurs.
Also soils aren't static, I've been on sites where the midpoint of a buried telecommunications cable had drifted over 8 feet from where the markers were a few hundred yards apart. Just looking at the markers and assuming a straight line isn't a safe bet. Looking at the fence rows to the left and right of the cable and you could see a matching bow in the fence rows.
I see you've never dug into a cable way out of range where the utility thought it was. You've definitely never been on a project where the utility was there with you wondering how the wire they marked many many feet away out of the hand dig area was actually where you're digging.
The UK ATC system was down again today for the second time this month. I'm no conspiracy theorist but hard to believe those two + this are 'accidents'. I believe the Netherlands has also had several issues recently.
So of course if the FAA is going to put AI into the watchtowers, they're obviously going to use local AI and not rely on some random Amazon cloud infra in some middle eastern country.
right. of course, no ones going to cut corners on the safe use of AI and secure infrastructure in this administration.
Pretty grim that a life critical system wasn't designed to report that the backup fibre was unserviceable until they attempted to switch over to it.
I wonder how long it was down? Days, weeks, months?
I am paying, $1000, $1800 & $1900 for the same service at 3 different location (20 mins from each other).
Two locations, I also have old coax lines that are still active, but not paying for it.
When I bought two businesses, I learned that they were paying for a dedicated fiber but using coax service.
At one of the location, we had fiber, and paying for backup coax and wireless. But if you turn off fiber box, it wont fail over to either one.
I wouldn't surprise it was down for weeks and no one bothered about it.
Over ten years ago, my employer was spinning up a new DC across the state and had three links between it and the primary DC. Two were pretty direct, but we needed a third because at one point in the 300 mile path, the two main links went within 400 meters of each other.
So how national-security-adjacent critical systems like an airport system doesn't have a larger set of backup links, and validates that they are geographically separate up until the connections, and have realtime alerting on the connection status, is surprising to me.
This isn't relevant in this case. The problems occurred at different places. The backup fiber was, separately from the issues with the primary system, cut by a construction crew. It's not two fiber lines both cut at the same spot.
Similar events happened in Europe disguised as thieves stealing fiber optic. This does not have any sense economically, as the value in the market is zero so... either is an honest accident and is cleared in a few days, or is sabotage
Do you honestly think that crackheads think that far in advance?
I've seen fiberoptic cables stolen from 2 (city) jobsites in the last 5 years, once by tweakers later caught trying to sell them as scrap copper and the second thief was never caught. This happened even though the spools had big signs on them saying "Fiber Optic Cable - NO COPPER".
I worked for a regional ISP that had a major outage when the redundant fiber provided by the telephone company was cut in one place triggering a full loss of connectivity. It also caused a massive 911 outage for 200,000 people as it isolated the 911 center from the city core.
Turns out the phone company didn't connect one side of the ring topology even though they certified they did. Needless to say lawsuits abounded.
The amount of shit I was given everytime I went through a checklist working at a fedramp certified company working with emergency alerts on the phone system made me prematurely gray.
I could feel the barely contained seething rage everytime I told the execs that the reason this release will take 3 days and not be instantaneous like your friends releases at a faang and that information made you embarrassed at your dinner party, is because you agreed to this process contractually years ago and now it’s a crime if I just sign off on it being ok without actually checking that it’s ok.
A lot of people are incompetent.
I worked at a company worth a few billion and the leadership balked when they told engineering they wanted a near instantaneous failover system and our department informed them that would require paying for a second environment that could be rolled over to.
It is rare to find leaders who can accept the cost of redundant infrastructure that is there for emergency backup.
What puzzles me is why they can’t accept it when they are perfectly fine with insurance costs and I can’t see much of a difference between the two when looking at a spreadsheet of costs other than possibly tax differences between the type of expenditure.
There's generally two wrong responses: (1) We spent a lot of money on 'blah blah blah', a lot of other companies use it, so yeah, we've got a backup/failure system. And, (2) inadequate testing - either, we tested 1 of 50 services, and it worked, so the whole system can be restored; or, we gracefully tested, and it worked, so it will obviously work during not-graceful incidents.
And the root cause of this is generally that no one gets promoted for implementing an adequate backup/failure system, or it's extremely rare.
Not saying that's what is going on here, just that it's possible.
Sometimes the level of incompetence / lack of care in organizations like this astounds me. I understand issues like this can be complicated and systemic but it honestly makes me think very poorly of the technologists building these systems in government.
BT guy ends that part of the presentation with "the IRA will really have to get shit together to take you off the net"
This was during the troubles. Same planet, different world.
Nothing I've ever seen or experienced with FAA indicates a lack of care; the parsimonious explanation is almost always that people who care a great deal are working within complex systems that don't always have a consistent or externally legible set of priorities. Or more intuitively, the failures we see are the "acceptable" ones versus the unacceptable ones (like planes falling out of the sky).
Hard to fathom someone still holding that view after the 737MAX disasters.
The NTSB has repeatedly and publicly criticized the FAA for failing to implement their recommendations, including as recently as last week (Amazon Prime crash resulting in 5 deaths). They've explicitly called out the FAA's failure to act as a contributing factor in crashes, multiple times.
Regulatory capture ruined the FAA.
https://www.airwaysmag.com/new-post/faa-smart-first-deployme...
The fundamental problem with DCA, and one of the two main causes of the crash [1], is that they are required by political pressure to operate at a higher operational tempo than they can safely operate at. DCA has essentially 1½ usable runaways--large planes can only use the larger runway, and given the capacity restrictions, airlines have been pushing to use fewer small planes at the airport. This shift in plane size means the effective safe slot capacity has gone down, but people still keep citing the same number as justification for safe numbers, and the politicization of the issue has shut down everyone who complained that the actual traffic just couldn't be safely handled.
If DCA's theoretical slot capacity (36 landings and takeoffs each per hour) is to be reached one just one runway, you have about 100 seconds to go from plane 1 touchdown; exit the runway, letting plane 2 on to take off; plane 2's wheel leaving the runway, clearing plane 3 touchdown. That's doable, but with essentially 0 margin for error. Alternating the runways used for each landing would give you closer to 30s of margin, but if only 10-20% of the planes can use the alternate runway, you can't divert enough planes to use the alternate runway.
Now DCA and the FAA aren't stupid enough to actually schedule 36 landing slots every hour, but the problem is that because of a few various factors, what was scheduled as, say, 30 slots for an hour (giving 120 seconds between landings, probably sufficient margin) ends up being 12 slots used in the first half-hour and 18 slots used in the second half-hour, which means a lot of the actual operation ends up having no safety margin even though on paper you have sufficient margin.
One of the ways you can rectify that is to assign slots further in advance so that you don't get the bunching. Effectively saying "oh, if you leave right now, you'll arrive at 3:30 with three other planes, but if I hold you for 10 minutes, I can push you into a less busy arrival time." Doing this requires good, accurate prediction of the actual flight travel times, and my understanding is that this is what the new software is meant to provide.
[1] The other main cause is essentially that the DoD's aviation practices in the area is a giant clusterfuck that endangers lives, and unfortunately that sentence is not relegated to the past tense.
Good luck to all of us.
The government politicians who approved this are frequent flyers out of there, so they’re at least getting their lines out on the line before the rest of the country if it proves ill managed or conceived.
It’s basically dogfooding.
https://www.cnn.com/2025/02/25/business/musk-faa-starlink-co...
Or is it that these ATC networks are their own air-gapped network with less redundancy? That just doesn't add up. Or maybe there was only one line going to the ATC, with no multiple "ISPs" like a datacenter would have?
* There is supposed to be a primary and a secondary link, in this case the primary failed and the fail-over also failed. It's unclear from the reporting if they were damaged in the same incident or if the failover was not tested or monitored adequately.
The Philadelphia TRACON site has been notoriously unreliable and was supposedly improved in 2025, it's also unclear if these issue actually could stem from that implementation.
First big oops of this form that I remember:
"In December 1986, the ARPANET had 7 dedicated trunk lines between NY and Boston, except that they all went through the same conduit -- which was accidentally cut by a backhoe. "
https://www.csl.sri.com/~neumann/insiderisks06.html
In any system where both sides of redundancy always carry traffic then loss of one link can cause congestion failure if any link fails. This is a very common means of failure in electrical networks that requires load shedding. Well, you can't load shed air traffic.
A more complex system that I like, but comes with it's own set of constraints and implementation issues is a system where both lines carry all the traffic at all times. This way the default state of the system is always working and your first failure isn't invisibly critical.
But this is very hard as we see in TCP when things get out of order and high latency creeps in. You have to manage a lot more state at the data level.
That seems like an extremely foolhardy thing to do.
It seems like it would present a less-chaotic solution than that provided by having no data communications at all.
[QUOTE:]
Russia has been staging covert operations and deniable provocations against European countries for many years. Since its full-scale invasion of Ukraine began in 2022 they have increased. These have included incursions by reconnaissance drones, hacking, arson, parcel bombs targeting cargo flights and plots to kill executives of leading European arms firms. In recent months the pace of such shadowy attacks appears to have risen again. The Russian operations have three main goals, according to analysts and security officials: trying to coerce European countries into halting their support for Ukraine; raising the costs of providing that support; and attempting to make NATO look powerless.
[/QUOTE]
And of course there've been "suspicious" cable cuts in the Baltic Sea and other European waters ....
https://www.economist.com/europe/2026/09/21/russias-grey-zon...
(The actual easiest method is to install the conduit or direct burial cable with tracer wire)
Before fiber a lot of things were put into the ground with no thought how you would find it again.
I sometimes hire directional boring and excavation contractors and if there’s any doubt as to where an electrical conduit or natural gas pipe is, I opt for the hydrovac truck to minimize risk.
And of course there are marker posts that say "buried fiber optic cable do not dig".
Most cheap hired labor doesn't realize you shouldn't dig up the market posts and throw them in the dumpster on construction sites. Add a little bit of rain and suddenly the contractor showing up with an excavator that was told everything was properly marked, and can see markings in other places but not where they are digging, and disaster occurs.
Also soils aren't static, I've been on sites where the midpoint of a buried telecommunications cable had drifted over 8 feet from where the markers were a few hundred yards apart. Just looking at the markers and assuming a straight line isn't a safe bet. Looking at the fence rows to the left and right of the cable and you could see a matching bow in the fence rows.
right. of course, no ones going to cut corners on the safe use of AI and secure infrastructure in this administration.