Inside Qantas.com's migration to the public cloud

Qantas expects to complete the massive shift of its core qantas.com website to the public cloud at the end of this year, following a multi-year effort that has seen the airline rework its approach to the migration.

In 2014, Qantas began exploring the possibility of shifting to a public cloud platform for qantas.com to better manage the million-odd bookings the website receives every day.

It spent 18 weeks trialling four public cloud services - AWS, Google, IBM and Microsoft - through three proof of concept trials: the build and deployment of a simple web application; the deployment of a legacy analytics application; and a "slice" of the qantas.com website, the search function for flight availability.

A year later, the migration began in earnest after the airline became "absolutely confident" in the model's stability.

The airline is now halfway through its shift of qantas.com to the AWS public cloud, having already moved 10 out of the website's 20 new microservices. It expects to complete the migration by the end of the year.

Updates that once took the airline months to deploy to the website now take just 30 minutes.

Qantas expects to save more than $30 million over five years thanks to the move, and has already experienced a 90 percent reduction in infrastructure costs.

Instigators of change

A year and a half ago, Qantas was left with little choice but to reconsider the entire stack supporting its core website.

"[Everything] was either near end-of-support of already out of support," manager of Qantas' applications centre of excellence Jessica Lin told iTnews on the sidelines of the recent AWS summit in Sydney.

"All the way from the servers to the operating system, the database, application server and even the content management system. That meant we were operating a very critical and important application for Qantas with lots of risk, and that includes potential security vulnerabilities as well as stability and resilience."

Not only would any outage of Qantas' core online operations cause damage to the brand, the revenue lost from any downtime would amount to $1 million for every hour offline.

A second driver for change was more strategic: consumer expectations with technological services are relentless, and Qantas needed to be able to push out new products and fixes to its website at a break-neck speed to keep up with demand.

"The challenge for us at the time was to put together a strategy that enabled and allowed our business to change quickly and innovate quickly, but in a safe and secure manner," Lin said.

The existing onsite infrastructure was also unable to cater to sudden changes in demand for scale, like when Qantas sales and promotions or unexpected weather events drive a large number of people to its site.

An additional desire to be able to "fail fast and pivot" with new technology led Qantas to look at how it could achieve all of its aims, and land on automation within a public cloud environment.

"We want to enable our business to get access to the technology and services they need through automation, whether it's giving them the access to compute power or server or storage at the click of a button, or to order a replacement of the app to spin up and test in a completely isolated environment," Lin said.

'[We also] wanted to be be able to achieve a lot of the things we do around security controls and compliance in the cloud through automation. Things like software patching, upgrades, scanning for vulnerabilities - we want to avoid human error and increase the speed at which we can deliver our software."

But first it had to demonstrate the value of shifting to a public cloud to the business.

Its second proof of concept, running its legacy flight planning application in AWS, proved that this setup could cut down the time it usually takes to calculate whether Qantas should operate in a particular market from three or four months, to just hours.

"It's really significant for the business to be able to see a critical decision that we can make in hours rather than months. And this was one way we could show our business that we have the technology at hand to do something that wasn't possible before," Lin said.

Costs and lift-and-shift

Qantas had initially gone into the project with the widely-held assumption that the biggest benefit from migrating to a public cloud environment would be the cost savings that arise from no longer needing to own and maintain infrastructure. 

It also assumed that it would be able to simply lift and shift its existing environment from onsite hardware into the Amazon cloud.

But as it embarked on intial pilots, it found neither to be the case.

While there was still significant cash to be saved, it was "icing on the cake" compared to what Qantas discovered were the real benefits: the speed and agility with which it could now jump on problems and push out new products.

Similarly, the team's initial plans to simply migrate qantas.com without much change to the application itself was never going to work, Lin said.

"We found out that if we continued down that path, we weren't going to get the most benefit out of the cloud," she said.

"So we changed our course. We started to rewrite the application from the ground up so we could make it more cloud-native, more of a microservices architecture."

Qantas decided to migrate each of these individual pieces in an iterative manner, starting with smaller, low-risk slices to help it understand how the tooling and software would operate in the cloud and give it more knowledge and confidence in the migration of the more critical pieces.

Taking the leap

Last November, the first low-risk slice of qantas.com, its weather application, went live in the public cloud.

A more critical application, Qantas' revenue-generating destination guides, followed soon after in December, as did its new content management platform.

A few more functional applications have successfully shifted over since late last year, and the team has been applying hot patches and smaller bug fixes in AWS at the same time.

"Our delivery cycle is a lot more frequent and iterative than we could have achieved before," Lin said.

So far the parts of qantas.com that have been migrated have experienced 100 percent uptime, Lin said - not because failure doesn't occur, but because the automation tools Qantas now has in place (like auto healing and auto scaling) mean the app can recover itself before someone "has to be woken up to respond to incidents like they did before". 

A dashboard tells Lin and her team how many services are running in the public cloud at any one time - last week Qantas had around 200 servers up and running to support more than 10 environments simultaneously for testing, pre-production staging and production, among other things.

"That's not something that was even possible or conceivable or financially feasible before," she said.

Lin's team will spend the remainder of 2016 migrating the final pieces of qantas.com into its new public cloud home.

"The challenge now becomes how do we keep up with the demands of the business and the expectations from consumers. Which is a good challenge to have," she said.

This project was named a finalist in the iTnews Benchmark Awards 2017.