Optimizing Costs and Delivering Images Faster Using Amazon CloudFront with Delivery Hero

Delivery Hero is among the world’s leading local delivery platforms, offering its services in around 70 countries across Asia, Europe, Latin America, the Middle East, and Africa. The company, which started as a food delivery service in 2011, now runs its own delivery platform on four continents. Delivery Hero is also pioneering quick commerce, the next generation of ecommerce, aiming to bring groceries and household goods to customers in less than 1 hour and often in 20–30 minutes.

Delivery Hero processes billions of requests and fulfills millions of orders daily. Customers place orders through a brand’s website or mobile app, where product images must load quickly to meet service standards. While Delivery Hero has long used AWS services, it was relying on Cloudflare to deliver these images.

In early 2024, Delivery Hero aimed to reduce costs and improve the performance of its image service. The company had earlier migrated that service from Google Cloud to AWS, and it wanted to consolidate its architecture on AWS. “The third-party CDN was an additional layer because we needed to transfer images out of AWS to the CDN and then from the CDN to the consumer,” says Wilson Tang, director of software engineering at Delivery Hero. “We wanted to simplify that process.”

After a successful proof of concept, the company decided to migrate its CDN to Amazon CloudFront, which securely delivers content with low latency and high transfer speeds. Delivery Hero wanted a speedy migration. Its contracts with the existing provider were coming to an end. A shorter migration would also help optimize costs because, during the migration, the company had to spend on both providers. It didn’t want interruptions in the user experience either, so virtually zero downtime was essential.

Delivery Hero discussed its sensitive migration with the AWS team and attended a workshop to analyze risks and understand the technology. The company’s three-engineer team realized that having a partner would speed up the migration, help the team learn more about using Amazon CloudFront, and validate the new architecture. The AWS team suggested tecRacer, an AWS Partner that had extensive experience with Amazon CloudFront migrations.

To onboard tecRacer, Delivery Hero used AWS Marketplace—where companies can discover, deploy, and manage software that runs on AWS. “Using AWS Marketplace, we simplified processes such as paperwork, procurement, and payment,” says Tang. The company saved several months on the process and could begin the migration sooner.

The Delivery Hero team planned and implemented the migration alongside AWS and tecRacer. The team refactored the architecture as well, and tecRacer used the AWS Well-Architected Tool—a trusted framework for evaluating cloud architecture and implementing designs—to follow best practices. “Collaboration is key for every project,” says Philip Lautenbacher, account manager at tecRacer. “This project was a perfect example.”

Delivery Hero uses AWS Enterprise Support, which lets organizations optimize cloud services. A feature of the service is access to AWS Countdown—which enterprises can use to optimize business-critical events, product launches, migrations, and modernizations on AWS. Using AWS Countdown for its migration, Delivery Hero worked alongside AWS to identify blind spots and potential performance bottlenecks, apply the necessary safeguards, and monitor progress to achieve operational excellence and avoid roadblocks.

Beginning in June 2024, the team migrated its image service in phases to Amazon CloudFront, using observability tools to monitor each phase. Working alongside AWS, Delivery Hero and tecRacer completed the migration in about 2 months—10 days ahead of schedule—with virtually zero downtime. “A project of this scale is challenging to do internally,” says Tang. “With the support and expertise of the AWS team and tecRacer, we migrated successfully.”

The microservices that manage the images run on Amazon Elastic Kubernetes Service (Amazon EKS), which organizations use to start, run, and scale Kubernetes. Delivery Hero stores more than 2 billion images, comprising petabytes of data, from its vendors on Amazon Simple Storage Service (Amazon S3)—object storage built to retrieve virtually any amount of data from anywhere. The company transfers images to customers through Amazon CloudFront and gathers metrics using Amazon CloudWatch, a service for observing and monitoring resources and applications.

Improving its observability with the new architecture, Delivery Hero can now make informed decisions and provide its internal stakeholders with more visibility into bandwidth usage. The simplified architecture has benefited all teams. “The outcome of the migration was amazing,” says Tang. “Our engineers are managing the CDN and other cloud services more effectively.”

By migrating to Amazon CloudFront, Delivery Hero reduced latency by 35 percent, speeding up the image loading time. Apart from improving performance and availability, the company reduced costs by 17 percent. “We use Amazon CloudFront for other services as well,” says Tang. “All of those services benefit from a lower unit cost.”

After its success with the migration, the team wants to explore machine learning on AWS to determine the suitability of images that its vendors upload and enhance their quality if needed. It also plans to explore options for improving the analytics that it provides to its internal stakeholders.

“This was a successful project,” says Tang. “There was virtually no business interruption, and stakeholders are happy.”