Qantas checks in with cloud computing

Qantas chief information officer Luc Hennekens has laid the foundations to shift many of the airline’s IT workloads to public cloud services, after being impressed with the results of three proof of concept trials.

The Qantas IT team spent an intensive 18 weeks trialling the capabilities of public cloud computing services offered by Amazon Web Services, Google, IBM and Microsoft to help build a case for a radically different approach to technology provision at the airline.

“The reality of cloud computing has taken over the initial hype,” Hennekens told iTnews.

“We’ve been talking about this for a decade but we are the point that it is working. The technology is not a bottleneck any more. It's tangible. It's very rich, it is at a great cost point and there is sufficient expertise in the market."

The outcome of the trials, revealed for the first time today, will have broad implications for a business looking to cut $2 billion from its costs over the next three years.

But for Hennekens and Qantas CTO Chris Taylor, the cloud trials have presented a bigger prize than reduced costs - a "tenfold" step-change in terms of the pace at which the business can deploy new services.

The two executives told iTnews the airline’s newfound competence in using public clouds will fundamentally change the way in which Qantas IT and its outsourced partners provision technology.

The three trials

The first of Qantas’ proof of concept trials was the build and deployment of a simple, three-tiered web application, standard fare for what organisations have deployed in public clouds to date.

“It was really just a trial of the commercial benefit;, the set-up, the configuration, and how we would want our vendor partners to run it,” Taylor said. “And what we found was that public cloud is now as easily consumable as it's purported to be.”

The second trial was somewhat more substantial - the deployment of a legacy analytics application to a public cloud.

For many years, Qantas’ performance engineering division has run complex calculations that take into account weather modelling and aircraft performance to determine the most fuel-efficient routes Qantas aircraft should take between any two points on its comprehensive global network.

The exercise is crucial to Qantas because fuel is a high proportion of the airline’s cost base, and better routing also helps streamline the scheduling of the airline’s fleet.

“This application runs flight paths against large volumes of meteorological and historical data,” Hennekens said. “The demand of compute power is absolutely massive.”

Running the route optimisation exercise for a single aircraft for five years would typically have taken Qantas four weeks on its own hardware.

For this trial, the Qantas IT team took advantage of ‘spot pricing’ on a unnamed but 'well-known' public computing provider.

Spot pricing enables customers to bid in an online auction for distressed inventory - unused server instances that had previously been ‘reserved’ by other customers who had later opted to sell access to them at a reduced rate.

The Qantas IT team won an auction for a large share of computing power and sent its modelling application and a huge sum of data to be executed on these server instances.

After a few minor edits to the application for deployment on a distributed x86 cloud, the team were able to calculate the optimal routes in under one hour. And to save on network costs, only the results were sent back rather than the full data set, the remainder destroyed.

For Taylor, a task completed in an hour that once took four weeks proved that the cloud offered features of “significant and material commercial benefit” to Qantas over the full lifecycle of an application.

“We found it was easy to install, easy to configure and easy to run,” he said.

“We auto-scaled it, de-scaled it,  and once compute was run we stopped paying for it - we only paid for what we used.”

By this stage, the IT team was confident in their competence with the necessary scripting and cloud orchestration tools and well grounded in the features and limitations of each of the chosen platforms.

The third proof of concept looked to win the hearts and minds of stakeholders elsewhere in the business.

The IT team took an undisclosed ‘vertical slice’ of the Qantas.com site - (a vertical slice might be the flight status page, the flight bookings engine or the frequent flyer program, for example) and deployed the production application in its entirety to a public cloud.

This required the swapping out of proprietary middleware the application had run on with license-aware, open source versions, the feeding of application source code into a hosted code control system and the spinning up of the necessary services from the public cloud vendor.

The installation and execution of the application was then scripted as it would be in a production environment, with precisely the same security and availability features coupled with it as if it had been deployed on Qantas tin. But now that it was based on a repeatable cloud platform, admins could push any changes to the feature live, with auto-scaling enabled, at the click of a button.

And while Qantas chose not to send the cloud-hosted version of the application live to customers, the team nonetheless threw 100,000 hits at it in under three minutes to test whether it would be robust enough for real-world use.

It passed the test.

“We’ve proved it,” Taylor said. “We now understand how we would port Qantas.com to the cloud. It's an excellent showcase to win hearts and minds. It's real and tangible.

“To build new hardware for Qantas.com alone took months in the past. And now, for the full lifecycle of the application, we can press the start button, and you can watch it all happen before your eyes, and it takes no more than 40 or 50 minutes."

Seeing the potential

The speed with which the team can now prove they can spin up new applications or iterative changes to applications has massive implications for Qantas, well beyond IT operations.

“Using the cloud accelerates the rate of agility - the rate of responsiveness for your business,” Taylor said.

If the airline was to conduct a large above-the-line, digital or media-driven marketing campaign, for example, the IT team would typically have required a minimum of two to three week’s notice to build the necessary infrastructure that would support the rush of potential customers likely to hit Qantas’ online platforms.

If new hardware is required, that lead time might have stretched out to months.

Once Qantas’ cloud strategy is played out, Taylor estimates the airline’s technology function might be able to respond to changes in the market within 24-28 hours, with new developments deployed with one click.

“Let's say we know we are going to make a change to the frequent flyer program, for example,” Hennekens said.

“We know that as a result, people will be logging into the frequent flyer calculator. Now you can add capacity to that function in a simple, reliable and cheap way.”

The IT team can now “spend more time thinking about the business of being an airline” rather than being engineering experts in every technology platform," he said.

The team will shift its focus to the tooling required to match the pace of cloud deployments, Taylor said.

“The old, traditional, labour-intensive, error-prone, housekeeping and boring activities you had to do in IT - things like managing license and hosting inventory - becomes automated in real-time using cloud platforms,” he said. “The cloud demands that you automate these things."

Public cloud is likely to be the default option for any new Qantas applications, especially web-based ones, as the airline also looks to port legacy apps over the longer term.

Read on to find out how Qantas’ cloud strategy will impact its relationship with suppliers...