by Thomas Macaulay

Trainline CTO Mark Holt on the company’s journey from cloud-first to 100% cloud with Amazon Web Services

Nov 22, 2016
Cloud ComputingData CenterRetail Industry

Trainline CTO Mark Holt believes the ticket retailer’s push to go 100% cloud has helped the company to avoid the delays that plague some of the rail franchises it serves.

“We are about high volumes, we’re about high transactions,” says Holt, whose company completes £2.3 billion worth of credit card transactions per year and sells 100,000 tickets every day.

“The key thing for us is primarily agility and speed. Time to market is everything for us. How quickly can we get releases out the door?”

Europe’s leading independent train ticket retailer has consistently embraced innovation since it was founded in 1997, in both gadgetry and culture. The app it introduced in 2009 has been downloaded more than 9.4 million times, and the company has recently empowered workers with a shift to cross-functional, product-centric, agile teams.

[Read next: Trainline CTO says customer-focused teams is success ticket]

Holt’s decision to move away from all of the company’s legacy tech in favour of a complete infrastructure migration to the cloud through Amazon Web Services (AWS) was bold nonetheless.

The figures suggest the punt is already paying off. The company is saving £1.2 million in capital expenditure every year while operational expenditure remains flat.

Point of departure

The move to cloud had a natural starting point, Holt told attendees at the 2016 CIO Summit as he discussed the importance of people and skills in the push to go beyond cloud first to 100% cloud.

“We had a big old legacy data centre sitting in Rotherham, hundreds of servers managed by CapGemini. And we had the quite nice situation where they were about to bulldoze it.

“It was a nice burning platform as well. You know there’s nothing quite like a tractor rolling into your data centre to make you move things quite fast.”

And move fast they did. An aggressive timetable helped complete the entire migration within 14 months.

“We used to be a very slow organisation,” Holt admits. “We did one release every six weeks. We had basically monolithic applications.”

The success rate of these massive releases was inconsistent, and moving to the cloud offered the chance to replace them with far smaller components. Since the November 2015 move, Holt’s team has released 20,000 of them into the cloud, at a rate of over one hundred releases per week.

“Our production environment now consists of slightly more than 320 components, each of which can be deployed independently,” says Holt, who joined Trainline as CTO in March 2014.

His team multitenants the components thanks to an infrastructure layer that allows different versions of components to be deployed on different sets of servers. Such a tactic wouldn’t be possible under the previous data storage arrangements.

“Now we can, in a cloud environment,” says Holt. “Whereas in an owned data centre you typically run maybe 40% CPU and then you start to get a bit edgy, in cloud you can run up to 60-70% no problem at all.”

The team set an autumn target for being in continuous delivery of most components.

“One afternoon somewhere in May, somebody said, ‘by the way, everything is in continuous delivery’,” he recalls. “It was a fantastic moment.”

Service disruptions

Rail industry regulations apply sparingly to Trainline, and are primarily concerned with outputs rather than what happens under the covers. To ensure the requisite Payment Card Industry (PCI) level one compliance, Holt involved the auditor in the cloud migration plans from day one.

“He came back every three weeks and we told him what we’d done and what we were planning to and how it was going to work and he was part of the journey with us,” says Holt. “He was part of the journey, which I think was very helpful.”

The approach reaped rewards when the PCI audit was held. Data was being encrypted as required both at reset and every time it went between any servers, and all of the auditor’s requirements were fulfilled.

“Most of the constraints were cultural,” says CIO 100 member Holt. “It started out by people not actually believing that I was serious.”

Empowering workers and harnessing developer enthusiasm for the cloud migration were crucial steps to a successful uptake.

“They all want to be learning about how it looks and how it feels to be playing with cloud stuff,” says Holt. “After we got over the initial hurdles it was very much about just them permission to do stuff.”

His confidence in his staff was the result of assiduous recruitment. Holt wanted the right people with the right skills who would be there for the long haul. They included one of the most experienced cloud experts in the UK, who now runs the entire infrastructure.

“Just all of those mistakes he’s seen 15 organisations make, he just solves them for us,” says Holt. “Whatever you’re trying to do, bring in awesome people. They made such a difference.”

Getting the board on board

The executive support and understanding of technology from both the Trainline board and the KKR & Co. L.P. private equity firm that backs it was a major boost to the move.

KKR acquired the company from buyout firm Exponent for an undisclosed fee in January 2015. What they want, says Holt, is an organisation that has already seamlessly integrated technology into the business. They also want a business that they can sell at some point.

“The nature of the speed improvements that we knew we would be able to get through moving to Amazon, our ability to just get [software] out the door, was really the key driver for it and they recognised that as a competitive advantage,” he says.

Holt is dismissive the concerns about cloud security held by other executives. “Is our funny little data centre in Rotherham next to the steelworks more secure than Amazon’s Tier 3 ½ data centre or multiple data centres?” he asks. “That actually is something they care about very deeply.

End of the line

Holt is an ardent advocate for AWS and other hyper-scale providers such as Microsofts and the Salesforces.

“They’re managing this massive infrastructure at hyper-scale, and the costs of them doing awesome security is tiny on a per customer basis, but it’s incredibly valuable for them to do it,” he says. “Go and talk to some of the other vendors and it’s basically just an old style data centre that happens to be in the cloud. That’s not quite the same.”

Holt believes that putting a product through a competitor’s environment shouldn’t be a barrier if you’re denying yourself access to the best cloud environment.

“If you’re competing on your ability to run infrastructure, then you’ve really lost already,” he says. “Compete on your user experiences, compete on top-line, compete on speed, compete on all sorts of things. But don’t compete on the ability to run services,” he suggests.

“Do you want to be on Amazon’s innovation curve where they’re doing 400 releases a year of actually really cool stuff, most of which is really valuable? Or do you want to be on some legacy vendor who is talking to you about VMware stuff, and you ask for a new server and it takes a really long time for them to install things?”

For Holt, the choice was an easy one. Trainline still can’t run British trains, but its move to the cloud has certainly smoothened the journey.