Total cloud: ClearScore eyes cost savings as it migrates everything to AWS

The London-based fintech has switched to AWS from IBM in an effort to cut costs and accelerate global expansion efforts.

AWS logo
Amazon Web Services

Founded in London in 2015 to offer customers free access to their credit scores, ClearScore has been shifting to a cloud native operating model since 2017, and is now looking to go one step further into the world of serverless computing to increase developer efficiency as the company’s global footprint grows.

ClearScore’s core credit scoring application pulls in data from credit agency Equifax and presents it back to customers, generating a commission if customers take out a product like a credit card off the back of their report. The application started life as a monolithic Angular application on a single Java API, hosted in the data centre of ClearScore’s biggest investor, Blenheim Chalcot.

At first, when the company needed to scale, it shifted the application to IBM SoftLayer (now known simply as IBM Cloud), while retaining the instance in the data centre to yield a hybrid cloud deployment.

Then, in July 2016, the UK’s Financial Conduct Authority issued new guidance for usage of outsourced cloud solutions in the financial services sector, allowing the credit scoring firm to start thinking about shifting entirely to the public cloud, to keep up with its rapid global scaling efforts.

By the end of the following month, ClearScore had migrated ten workloads to the public cloud with Amazon Web Services (AWS). By April 2017, the company had decided to go all-in. Klaus Thorup, CTO at ClearScore, cites the wealth of AWS experience in the jobs market as a key factor in choosing AWS as its sole infrastructure provider. AWS’ global footprint was another attraction, as ClearScore looks to expand services into Africa, India and Australia in the coming years.

But the move couldn’t occur without some serious preparation. "A lot was about setting up the pipelines and the re-architecture of the API to allow for a move to EC2 and making sure we had the primary Postgres database switched to a cloud-oriented version," Thorup explains.

The database was the first thing Thorup and his team shifted to AWS. Then a dedicated team moved microservices over one at a time until they had feature parity with the old IBM-hosted environment. A 12-hour cutover later, and the company was all-in on AWS.

Were there fears of vendor lock in? "[It's] not a concern," he said. "The ecosystem expands so rapidly and there are a lot of various services we can take advantage of. That being said, if things do change and someone offers a better service, it may be a hybrid approach that we consider. But at the moment we aren't concerned by the stability and offering of AWS."

"It has been a journey"

Like any cloud-native company worth its salt, ClearScore is managing these microservices in containers. "It has been a journey," Thorup said, stifling a laugh. "It has been an evolution rather than revolution."

Engineers had to shift from its previous HashiCorp stack to AWS, move from the Nomad scheduler to Kubernetes, and quickly build up skills over the past couple of years through internal training dubbed “ClearScore University.”

"There is a high learning curve, but we have found the market has embraced these technologies. So when it comes to hiring, if they don't have the skills, there has to be an appetite to learn," Thorup said.

The results so far have been an uptick in application runtime, from 99% to 99.97% – just short of the holy grail of five 9's – and a release cycle that works in hours not days.

Release velocity is a "north star metric" at ClearScore, allowing engineers to operate autonomously and providing users with 5,400 updates and feature releases to products per month, according to Thorup.

For example, ClearScore recently released a new feature called Resolve, which provides customers who experience financial difficulties with suggested methods for debt consolidation.

What next?

Now that the company is firmly in the public cloud it is starting to look forward to automating away more operations tasks and eventually going as 'serverless' as possible. Engineers have been getting to grips with Lambda triggers "for newer parts of the platform to segment and do greater personalisation of users," Thorup said.

"We use Lambdas predominantly around newer parts of the site, for example with real time data segmentation of users and to power our partner portal service to access data around the way we utilise their services. A lot of that traffic is not consistent or can be at a low level at times, so for ease of architecture Lambdas allow us to quickly scale up," he added.

ClearScore is also interested in using more managed services to ease the operations burden around containers and Kubernetes, so AWS EKS is likely on its radar. The company also wants to start using more spot instances – where AWS sells unused server capacity for a cut price but can take away those nodes as and when they need them back – to further drive down costs by 25 percent, which would require some flexibility and internal expertise to do.

Eventually, ClearScore wants to run its technology function on a NoOps model, "where the majority of infrastructure is looked after and we have a ten to one dev to ops ratio, so squads can independently release," Thorup explained, allowing the company to keep up with its ambitious global expansion targets, and continue to release at high volumes.


Copyright © 2020 IDG Communications, Inc.

How to choose a low-code development platform