Mainframes are inherently broken. I'm well aware that they work and are actually quite efficient, however they carry limitations that would end them at any company not run by retarded boomers:
1) They're monolithic systems, which makes it severely problematic to split off or upgrade portions of them. Which is a huge problem for 24/7 operations and technical debt.
2) Batch-based processing is very efficient, and a complete shitshow once job sizes exceed normal volume because there's no quick or reasonable way to scale up capacity. Exhibit A: the Southwest disaster.
3) Incapable of Active-Active High Availability. Mainframes require someone to declare a disaster (which boomers are too scared to do because their shitty ancient systems dont reliably come up from a standby data center). You cant simulate real world problems by randomly yanking cables or shutting down processes Chaos Monkey style without the shitty old mainframes falling over like a boomer with myocarditis.
4) About as dynamic as the tax code. Great if you can manually or batch import all your schedules months in advance, but if it falls over you have no sane way to get schedules running again today. VS a more modern system like Uber where "where are our drivers and vehicles so we can get trips scheduled and seats filled" would be relatively straightforward after an outage.
2, 3, 4 are on point. While you are partially correct on #1, airlines have zero choice but to use mainframes. Nothing else is going to keep up with those levels of transaction processing. If what the unions are saying is true and nothing has been upgraded since the 90s, it's likely SWA has 390s or early z systems. But that would mean that SWA is either running unsupported or they are paying a fuckton in extended support.
I am willing to go out on a limb and say the unions are full of shit. Just because an interface hasn't changed much in 30 years doesn't mean the backend systems haven't. User transparency and experience consistency is in the IT bible. I had to explain that once to a "senior infrastructure and design" manager. He was trying like hell to get rid of a (recently purchased) mainframe because it he swallowed M$ legacy spiel like a fucking fish and was being reeled in by his asshole. People couldn't learn it but they could learn a pretty gui. Guy was a dumbfuck. When that company got a new CEO, he was canned and walked out with the entire C-level and upper management. He's probably ruin lives elsewhere now.
Anyway, back to SWA (and I really hope someone that works there is a poaler) they need (STEP 1) to get BCDR people on staff ASAP to build out an active-active failover structure. I have seen it done and designed it before with z-series so I know it can be done. Basically, push button failover. Automate the fucker so good that a button is pushed and bam, operating out of another data center. (STEP 2) is more difficult. Get rid of all the checkbox hires.
(post is archived)