bloomicon -

Atom Bank CTO on ditching its datacentres and moving to the Google Cloud Platform

Atom Bank CTO Rana Bhattacharya describes what is behind the mobile-only challenger bank’s large-scale, cloud-focused reboot of its infrastructure setup

Atom Bank CTO Rana Bhattacharya has revealed how the next phase of the company’s growth will be underpinned by its increasing use of the Google Cloud Platform (GCP), as part of ongoing work to downsize its dependence on third-party datacentres.

The mobile-only challenger bank started out in 2015 with the self-appointed goal of becoming the UK’s most customer-centric banking institution.

At the time, the Financial Conduct Authority’s guidance on the use of cloud by banks  was still being hammered out, so Atom Bank had little choice but to house its infrastructure within third-party datacentres.

“We were on a journey from 2015 to get a bank live and, at that time, the regulators’ views on using cloud were pretty immature… so we had to leverage partners that could provide us with infrastructure like datacentres,” Bhattacharya tells Computer Weekly.

Like all challenger banks, Atom had post-launch ambitions to not only expand the range of banking services it could provide, but also to increase its customer base, and soon the limitations of its datacentre reliance started to become all too apparent.

“We wanted to accelerate our velocity of change, and what we found was that if you wanted a new environment [under the datacentre model], you had to go through this whole process where you have to work out if you have space in a rack to put in more blades, then purchase the blades for the rack, install software and build the stack up on a new server – and that takes a lot of time,” he says.

At the same time, all this work has to be negotiated via a third-party datacentre operator, which complicates matters further from a procurement and contractual perspective.

“This isn’t an issue with the third party, it is an issue of the outsourced datacentre model,” says Bhattacharya, “in that you have to go through the steps to get what you want, which takes time and money, because you’re paying multiple people to coordinate it all.”

All change on cloud

Within a year of Atom Bank’s launch, the FCA issued its guidance to financial services firms about how they should go about outsourcing their IT to the cloud and, in doing so, the path was laid for the firm to start switching up its hosting arrangements.

“We knew the growth we wanted to make happen and the pace we wanted to bring services to our customers, and we also knew the traditional datacentre model wouldn't support that,” says Bhattacharya. “And at the same time, we wanted to be in control ourselves.”

Atom’s push to get out of the datacentre has seen the firm publicly commit to using fintech firm Thought Machine’s cloud-based banking platform, Vault, to record and manage transactions made within the personal and business banking products it is bringing to market.

Alongside that effort, Atom embarked on a request for proposals (RFP) for additional support in its move off-premise, which Bhattacharya says garnered responses from “all the typical players”, before it whittled down its choices to two cloud firms, with Google emerging as the outright winner in late 2017.

“We wanted someone that would truly be a partner with us on our journey, as 50% of the battle is actually having someone on your side to make you successful,” says Bhattacharya, who joined Atom Bank as CTO in April 2017.

And with Google and its Thought Machine technology tie-up in place, Atom Bank has set about re-architecting its entire setup and processes to ready itself for a wholesale move off-premise.

“What we have done is not necessarily just a lift and shift model move of what we have in the datacentre to the cloud, and we have used this as an opportunity to basically rearchitect our future bank,” he says.

Ultimately, the company wants to have more in-house control over its IT, which it has set about achieving by adopting DevOps software development processes and creation of what Bhattacharya calls a “slick, high-end” site reliability engineering (SRE) function.

“This means that we can, through our own controls, effectively spin up a whole version of the bank, as we need to,” he says. “We wanted that ability and control to be in-house. So that it is our bank, and we have that control.

“Google and our DevOps capability are the foundational bones [for Atom] to basically integrate with a variety of other third parties and software-as-a-service [tools] as well as hosting our own components that we will build on the fly, and test and deploy as needed.”

Spreading the work out

There are about 90 people within the Atom Bank tech team undertaking all this work. They are not only tasked with overseeing the cloud-focused update of its banking platforms, but also ensuring that these offerings integrate with the older parts of its infrastructure.

“That is the team that runs the existing, current bank, as well as all the change as we work towards creating our future bank,” he says. “What we’ve done is essentially build a new bank on the side of our existing bank.”

In terms of where the organisation is now with its move to the cloud, Atom Bank is just about to go live with its new consumer-facing banking app, which will be hosted on the Google Cloud Platform, as will “a good number” of IT workloads, says Bhattacharya.

Next on the agenda will be the launch of its core banking platform, whose development has been shaped through its partnership with Thought Machine, which will also go live on Google in due course.

“We are trying to switch on different components as they are ready, and in a very safe way, incrementally, and add functionality to that as we go to allow us to accelerate our journey,” he says.

“Our focus is on getting live, and doing that in a very economical way, and optimising what we have built as well.”

Read more about cloud and financial services

And once most of its workloads are on Google, the next thing on Bhattacharya’s tech to-do list will be to go even deeper in to the cloud giant’s extended service portfolio.

“Once we’ve got the majority of the bank up and running on Google, then we can actually do some really cool fun stuff and leverage the full capabilities and services, because they are doing a lot of stuff that is above and beyond just the pure infrastructure,” he says.

Machine learning and real-time data analytics are a couple of areas Bhattacharya has earmarked for experimentation later down the line.

For example, he describes how combining the Thought Machine platform with the Apache Kafka and Tensorflow on Google Cloud could generate real-time data feeds on how various parts of its bank are performing.  

“We could have real-time analytics running to give us a view at that point in time to the second about the aggregate balance across a particular product, and build an understanding of our customer journeys and what elements we might need to focus on to improve,” he says. “So, over time, we will be leveraging data with machine learning to do some really cool fun stuff for our customers.”

Read more on Software-as-a-Service (SaaS)

Data Center
Data Management