Visa outage highlights IT maintenance challenges – and the promise of predictability

In June, Visa started rejecting one in 10 financial transactions across the U.K. and Europe – a problem lasting 10 hours and affecting 1.7 million cardholders. Even in an IT environment designed to support 24,000 transactions per second, a hardware failure crashed the system. The incident was a wake-up call for an industry reluctant to suspend services for scheduled, expensive repairs. Could predictive maintenance have prevented the crisis?

Predictive maintenance draws on machine learning, neural networking, and artificial intelligence. Commonly used in marketing, learning technologies improve with use: every time you search Google, its accuracy improves.

Yet while AI can predict preference, it is still learning how to factor in context. Nirvana for marketers will be when technology shows my car purchase is followed by a caffeine urge, with my coffee advertised accordingly. It’s the search for the unforeseeable yet real relationship that can only be found with a deep data dive. We’re not there yet, but we’re on the way.

Maintenance that informs itself

The same neural networking technologies are being applied to hardware and networks. There is countless data in a data centre. Just as marketers want to utilise all the information available, so do data centre managers. The promise in machine learning is the ability to examine the full range of performance data in real-time to detect patterns indicative of “faults-in-the-making”, uncovering relationships no human engineer would return, like cars and caffeine.

This application of AI algorithms to data centre maintenance underpins our ParkView advanced monitoring system, which contextualises patterns to “understand” infrastructure behaviours. This means instant fault identification and fewer false alarms. Future predictive systems will prevent the types of issues Visa experienced.

The next stage: predictive maintenance taps IoT

In the Tom Cruise sci-fi movie, Minority Report, police use “psychic technology” to prevent crimes before they happen. The twist comes when the crime-solver is accused of the future murder of a man he hasn’t yet met.

There is a parallel with data centres. Human error causes an estimated 75 percent of downtime. That’s why data centres are less populated. The perimeter has security staff, but the interiors are becoming vast and lonely server expanses, where the electric hum is rarely broken by the sound of footsteps. The downside is the lack of human detection of things like temperature changes and dripping water.

That’s where the IoT and the Industry 4.0 playbook developed in heavy industry comes in, in which remote monitoring enables smart and predictive maintenance. A good example here is fixing a data centre air-conditioning system based on its predicted performance in relation to it’s surrounding environment. This concept can be applied across the entirety of a data centre and its cooling, power, networking, compute, storage, and other equipment. Emerging dynamic and largely automated predictive maintenance management will transform the data centres we know today into self-monitoring, self-healing technology hubs, enabling reliability as we move computers to the edge to support the IoT applications of tomorrow.

Evidence indicates a move from a reactive/corrective stance, still dominant in many data centres, to more preventative maintenance delivering average savings of up to 18%. The next leap towards predictive maintenance drops spending about 12% further. In fact, Google used such strategies to drive a 15% drop in overall energy overhead.

Combating downtime with predictive technology

Enterprises must integrate predictive maintenance. Downtime kills reputations, profits, and customer relationships. Most organisations like Visa can recover from unplanned outages, but reducing unscheduled maintenance is always preferable.

IT leaders must make hardware and facilities as downtime-proof as possible. This means using machine learning and AI to return a pound of ROI on every ounce of prevention possible. Banks are investing in AI for a range of purposes, from contract scanning to fighting fraud. It’s essential that the new technology is used to fix problems in advance.

By Evan Kenty, Managing Director EMEA, Park Place Technologies