> its not shocking to state they are a very ancient infrastructure that is being run and protected by fiefdoms that refuse to even acknowledge best practices of any infrastructure tech released in the past decade
I struggle with the notion that a high quality airline operating system cannot be developed using technologies as of 2015. Most of what we are drowning in right now is the product of the last 10 years.
The last place we need fancy new shit is in air travel. This is precisely the kind of thing where you do want to call someone like IBM to install a mainframe. Failure of an airline's IT systems can begin to approach the kind of impact you get with a payment network outage.
Hell, you could run a high quality airline on the tech of the 60's. You could run a high quality airline on the tech of the 30's nothing except radios and the planes.
It's not a tech problem, it's a culture problem. Just because the infrastructure is old does not mean that it is bad. The main deciding factor is how well it is maintained. But that is to hard for many people. So much easier to say "It'S bAd bEcAuSe iT is oLd" and walk away.
Most are - irony of irony is that when Delta had its big outage due to CrowdStrike, dealtamatic/deltaterm was still running just fine, but no one could get into it because all of their windows machines were locked out.
At the core of most airlines is a customized version of IBM TPF, its very reliable and highly available, its all of the other stuff that breaks down.
We will in time find out what grounded AS, I wouldnt be surprised if its some sort of middleware connecting their iPads to the CRS they use for ticketing operations, but it could also be something as simple as their weight and balance application going offline.
AS is a fairly well run airline (as are DL, AA and UA) with a heterogeneous mix of systems in service, ideally this heterogeneous nature should make for a more resilient system but it also can lead to single points of failure when you have to glue too many different systems together.
excellent point. at tmobile (circa 2014) it required 87 APIs be hit to turn on a new subscriber. if the 34th API failed, the subscriber had to wait and start over clogging up the stores. at the core only 3 or 4 of those APIs were crucial to start the service and the rest could have been fine with eventual consistency. who is ticketed for what flight is the core, the rest dealing with plane can be handled manually just like the small airplanes do it, manual weight & balance, flight planning, etc. but Alaska chooses not to do that and is ok losing millions of dollars per day disrupted while losing customers because they do not care. and not hiring a safety officer for years proves they do not care.
I am by no means an industry insider, but I’m skeptical of your claims about running a ln airline on tech from those eras. The visible side of airline IT (ticketing) perhaps, but surely there is a lot of behind the scenes software that facilitates the efficiency of operation (plane positioning, route planning, maintenance tracking) required to compete on price in the modern era.
It’s easy to complain about modern airlines (and I do), but it’s still true that’s never been cheaper to fly, and IT infrastructure is surely no small part of that.
I think the ticketing systems are probably the most modern parts of airlines. As far as I know, the tech that actually runs the plane does not change very often as it needs to go through approval processes.
The ticketing system might very well be the oldest.
AFAIK the very first large-scale commercial deployment of what we now call "distributed cloud apps" was SABRE, a ticket reservation system built back in 1960s, still in use today.
that was my point in the purposeful use of the word "fiefdom" to describe the Alaska IT culture. it wasn't focused on optimal infra state with what it had, it was focused on following the cult of the winders wizards that refused to acknowledge anything different like Linux.
The US has a severe brain drain issue in the tech industry. American companies don't pay good salaries for people who specialize in well tested technologies, like the ones you mention from 2015. These same companies prefer to throw tons of money at shiny new things, like blockchain, AI, or whatever the next buzzword will be. Engineers in stablished tech areas will either have to move with the crowd or retire, and never be replaced. New engineers will by necessity have to learn the new shining tech. So the answer is that, yes, we could do these things with 2015 tech, but we cannot because they won't pay experienced people to do this.
It can be developed. I think it’s more of an indictment of the type of company that doesn’t even consider changing anything not currently screaming bloody murder at them.
I struggle with the notion that a high quality airline operating system cannot be developed using technologies as of 2015. Most of what we are drowning in right now is the product of the last 10 years.
The last place we need fancy new shit is in air travel. This is precisely the kind of thing where you do want to call someone like IBM to install a mainframe. Failure of an airline's IT systems can begin to approach the kind of impact you get with a payment network outage.