British fashion and beauty ecommerce giant ASOS is in the middle of an ambitious transformation of its entire technology stack, shifting to Microsoft's cloud Azure platform-as-a-service (PaaS) and a host of software-as-a-service (SaaS) solutions for its backend enterprise systems.
Speaking at the MuleSoft Summit in London last week, ASOS CTO Bob Strudwick ran through the transformation the company has already completed, and what comes next.
"The transformation programme started four years ago with the decision to replace everything," he said. "The software stack which the business had built had reached the end of its natural life and had become increasingly difficult to modify and scale."
The transformation hinged on three main targets, according to Strudwick. "We needed an architecture which provided a globally consistent performance and high availability," he said. "We needed the flexibility to introduce new features and we needed the ability to present those new features in innovative ways for our early adopter 20-somethings."
Everything is then underpinned by a big data platform which has to be able to handle 15 million regular users - which ASOS calls active customers - shopping across a variety of channels.
The data lake itself now comprises multiple Microsoft Azure products: namely Azure Data Lake Store for long-term storage; SQL Azure, Azure Data Warehouse, CosmosDB and Azure Data Factory for distributing data.
On top of this, ASOS wanted to be able to leverage data science to get insight from this data, specifically using the Scala/Spark MLlib machine learning library on Azure HDInsight for product recommendations, and the open source machine learning framework TensorFlow.
ASOS, naturally, has shifted to a microservices architecture, now all running on Microsoft's Azure platform-as-a-service.
This included shifting key consumer-facing applications, including the mobile apps, to be hosted on and interact with the Azure PaaS, which was all completed in time for Black Friday in 2016. Next, ASOS is redesigning and rebuilding its web application using React JS and Node technology ahead of this year's busy shopping day.
Once these frontend services have been rearchitected and shifted, ASOS will focus on its enterprise software for the next financial year. The new stack will look roughly like this: Microsoft Dynamics 365 for finance, Workday for HR, Oracle RMS for retail, Stibo Step for product information management (PIM), and ADAM as its content hub.
"Right now the new digital platform coexists with legacy backend systems through a service tier and one by one these financial, retail, people systems will be replaced incrementally," Strudwick said. "So we have some work ahead to be sure."
Integration Platform Team
ASOS keeps everything integrated during this process through a central integration platform team, whose job is to act as a centre of excellence. This team engages with the business to define standards, best practices and patterns, build common components and design and build MuleSoft environments.
"In our world the integration platform team is about setting standards of how we expect the technology to be used and then they work with teams across retail and finance, for example, to do the first interface end-to-end to establish patterns and how they set up continuous integration and deployment, so establishing an engineering approach," Strudwick said.
"My view of it is that building the foundations first and that integration platform team needs to be outward facing to help drive that cultural change and be focused on helping those platforms in finance and retail etc to adopt new technology.
"In our legacy stack we had every integration technology known to man so this is a big cultural leap for us."
Finally, on MuleSoft specifically, Strudwick said: "For us the decision turned on that we liked MuleSoft as an integration-focused organisation with momentum in the market and I guess the IPO has cemented that idea that we were right to back them.
"We had used MuleSoft in the ASOS China stack so we knew it was ready for prime time and lastly we liked the Azure story and integration."