Back in 2007, I wrote a blog post that complained about an outage at United Airlines that affected me personally. I suggested several ways for the airline to improve its IT architecture -- it looks like United didn’t listen.
Last Wednesday, United's management system once again took a dirt nap. It knocked the airline's computer system offline, according to media reports and Twitter posts. Passengers were unable to check in for flights, and United couldn't print boarding passes or luggage tags. Thankfully, the glitch didn't affect the flight plans of planes already in the air, but United pointed to a network issue that brought the airline to its knees.
The rule is that you design systems so that no single component failure will bring them down in their entirety. If you can’t figure out how to do that in your enterprise, then for God’s sake put your systems in the cloud to gain the resiliency you need.
United suffers from the same problem as many enterprises: the inability to create a fault-tolerant infrastructure. Public clouds such as those from Amazon Web Services, Google, IBM, and Microsoft are not perfect, but they seem to keep systems running around network failures, power failures, and even human mistakes.
My advice to United: Take a port-and-improve approach. When migrating to the cloud, look for opportunities to make the systems automatically work around hardware and software failures. A single router, server, or network connection failure should not be able to stop production. Test these workarounds often to ensure that what happened last week at United does not happen again.
If United is unwilling to invest in making its own infrastructure resilient or it doesn’t have the money to fix its IT problems, then migrating to the cloud might be a good idea. Of course, that will mean United’s IT department can’t hug its servers, but giving up control is part of the deal.
If United does neither, more outages will follow -- and more customers will fly elsewhere on airlines whose planes can actually get off the ground.