Qantas' cloud-based flight sim saving millions in fuel

A cloud-based flight path simulator, which determines the most efficient route airplanes can take to reach their destination, is set to save Qantas millions of dollars in fuel, the airline's CEO Alan Joyce said today.

The Constellation system simulates thousands of possible flight paths based on "millions of data points" including weather and wind patterns, and selects the route that uses the least fuel.

"From a business point of view, that's going to save us $40m in costs each year by a one or two per cent improvement on these flight plans," Joyce told an audience in Sydney this morning.

"We did a Sydney to Santiago flight – we saved one tonne of fuel on that flight by getting a better flight path. And we're going to reduce our carbon emissions because of that system by around 50 million kilograms each year," he added.

The "numbers are massive" and will have a "huge impact on our business" Joyce said at Amazon Web Services' Sydney conference.

The system, which Qantas group chief technology officer Rob James described as a "new flight planning algorithm", was developed in partnership with the Australian Centre for Field Robotics at the University of Sydney.

Constellation, understood to have been around five years in the making, will be rolled out to all aircraft by the end of the year.

Qantas' fuel bill for last year was $3.2 billion, making a one per cent saving hugely significant.

"One of the most expensive parts of flying is fuel. It's also the most volatile cost to an airline. Anywhere between 20 and 40 per cent of an operating cost for an airline...Every time one of those fully loaded A380s take off we are pumping about seven litres into those engines every second," James said.

"That's like a fire hose of fuel going into those engines. What you want to do is fly those things as efficiently as possible," he added.

The system – which runs on AWS – sometimes picks routes a human operator would not, for example going miles off course to pick up a time-saving tailwind.

"We can choose to take a less turbulent path, or we can pick up a tailwind to get you to your destination much sooner," James said.

Flying towards the future

James also revealed the airline was using another simulator engine – dubbed QuadraX – which simulates how aircraft will perform in different conditions and configurations over a 15 year period.

"Apart from fuel being so expensive, the aircraft themselves are very expensive assets to purchase. And when you buy one that's a ten to 15 year commitment so you want to get that right," James said.

QuadraX allows airplane purchasers to simulate the assets' performance on different flight paths, with different engine configurations and usage patterns and in different weather and traffic conditions.

"Every time we run one of these simulations we're basically doing a ten year study in about nine hours," James said.

The analysis is compute heavy and done in the AWS cloud. 

"We spin up approximately 4,000 CPUs, run the analysis and then spin it right down to nothing," James said.

The CTO said cloud would become an even bigger element of Qantas' operations in coming years, driven by the increasing amount of data from the airline's assets.

"We've got these new Dreamliners, 787s, and every time they land we download half a terabyte of sensor data that we can analyse; every time they fly. And that is dwarfed in comparison to what we expect in the future. The next generation of aircraft are going to give us a petabyte of data every time they fly, from the engines alone," he said.

The company has a roadmap to move most compute workloads, including mainframes and mid-range systems to the cloud by 2021.

"A pretty bold goal for us," James said.