Call me a cynic, says IoT industry observer and freelance writer Nick Booth, but something about the recent British Airways computer meltdown doesn’t add up.
The airline has ordered an independent investigation into the systems failure that grounded 75,000 passengers during a public holiday.
You don’t have to be a data centre expert to be suspicious of boss Willie Walsh’s explanation. It was all the fault of a technician who switched the power supply off and on again, says Walsh.
Hang on. It’s hard enough to get inside if you’re a qualified engineer. It’s not possible that BA let a self-learner in the building, let alone give him licence to experiment.
Granted, there is an odd culture in data centres. Asking people for help is anathema. My boss in a similarly grim gulag that ran London’s Metropolitan Police communications centre always insisted he was self-taught, in some lame attempt at machismo.
His lack of social skills forced me to experiment with the system, to see how things worked. This had disastrous consequences for London’s traffic one busy Friday. It could have cost the police a fortune, if they hadn’t omitted to tell the public they were entitled to compensation.
To summarise, BA’s problem appears to be that interpersonal communications in its data centres are unacceptably amateurish, and the human capacity for self-learning is positively dangerous.
This, surely, is where machine-to-machine communications could prove its worth. Losing £150 million (€173 million) over miscommunication is “absurd” in the age design for failover, says Mark Skilton, professor of Practice in the Information Systems and Management Group at the UK’s Warwick Business School.
There are several areas where machine learning algorithms and artificial intelligence (AI) could have intervened before this rogue technician’s experiments brought the system down. Commission-hungry IoT sales staff should be bombarding Willie Walsh right now!
Data centre designers should use AI to test virtual model simulations of the real-world system, you could tell him. So even an unplanned failure can be neutralised if the design of the uninterruptible power supply (UPS) is checked and installed differently. Skilton is convinced BA’s failure was nothing to do with the UPS anyway.
Maybe the ‘power failure’ is the story they are telling BA’s shareholders or the insurance company. Who knows?
Before the insurers pay out and bump all our premiums up to compensate, maybe one of you should tell the loss adjustors what neural networks could have achieved. They could have spotted an anomaly (such as the presence of a techie with a manual) and created alerts. When he started experimenting with the power supply, surely someone’s screen, somewhere, should have lit up like a Christmas tree.
Artificial Intelligence can conduct smart record checking of personnel. The common sense of many security guards in data centres is often all you could expect for a disinterested man on a minimum wage. Yep, I’ve done that job too. My CV says I was “fired with enthusiasm”. During the BA event a blend of augmented reality and AI could have been used to check human engineering working procedures, says Skilton.
Many of the companies featured in IoT Now could train staff and subcontractors while automatically checking that processes run in the right sequence. Given the confusion created by mainstream media reports over the BA incident, explaining this could be a challenge. Skilton uses popular games in order to exemplify and popularise the topic for the layman. “Think of it as Pokémon Go with the engineering drawing overlaid on your eye ball through smart glasses or a mobile or tablet to assist checking,” says Skilton.
Robotic control on automation of critical switches could have been put in place. This is standard practice in the Gas and Nuclear industries, where flammable high-pressure pipes need to be monitored. Given that BA cannot function – and its staff can’t even use their common sense – without their IT system, one would have thought this would be a massive priority.
Ultimately, the automated failover to an alternative data centre and websites could have used better machine automation and decisions. And the aftermath of the event could have been automated too with AI managing the social media. A Twitterbot would have been more communicative with the disappointed passengers than many of the BA ground crew!
The author of this blog is Nick Booth, freelance IT and communications writer