Tim Rutte

The traffic attribution of a global AdTech platform was distributed across several departments – no central system, no uniform process. With increasing traffic and a growing number of campaigns, media buyers could only track a fraction of the possible campaigns.

The requirements were uncompromising: Enrich every request with attribution values in under 10ms, no single point of failure, globally available in EU and US. Assign attribution values on a daily basis, managed consistently and atomically.

Go as the language of choice for high concurrency and low latency. Deployment in EU and US, each region independently highly available. AWS Global Accelerator routes traffic to the nearest region, an ALB distributes the load within the region.

Three cache layers: In-memory for frequently requested values, Valkey for distributed caching, DynamoDB for persistent daily assignment with atomic transactions – no race conditions, no inconsistencies under load.

The service is multi-tenant capable from the ground up: multiple tenants isolated on the same infrastructure, without mutual interference.

A distributed manual process was replaced by a central, highly available service that automatically scales with traffic.

A newly formed team took over PHP legacy code for revenue reporting, grown over years: no test coverage, no documented infrastructure, no Infrastructure as Code. Revenue feeds from external partners were imported daily, campaign reports were delivered. Errors had immediate business consequences.

No tests means flying blind. Incidents happened regularly, causes were hard to find, fixes risky. The infrastructure was manually configured, undocumented, not reproducible. Task: modernize, migrate to PHP 8, transfer to a clean AWS environment during ongoing operations, no downtime allowed.

Built test coverage from 0 to over 90% – PHPUnit for unit tests, Codeception for acceptance tests. Secure first, then modernize. Migrated all legacy code to PHP 8.

Cloud migration to AWS: ECS, EC2, SQS/SNS, RDS, Athena, and Glue. Entire infrastructure in Terraform, versioned and reproducible. Estimates-Per-Click calculated daily and provided to media buyers. Revenue data made available internally across all departments via AWS DataZone.

Legacy modernization during ongoing operations in a nine-figure revenue environment, not a single day of downtime.

An international SaaS product with several million customers operated a payment API that was built for a different scale. Every load spike was a risk, new provider integrations took weeks, deployment was a manual process. The product was intended to scale internationally, but the system was not built for it.

Complete rewrite was not an option. The system processed thousands of transactions daily; a failure meant direct revenue loss. Modernization during ongoing operations, new functionality in parallel to the existing system, over several years without destabilization.

Introduced test coverage, documented critical dependencies, automated deployment. Then gradually implemented a microservice architecture in Go, no big bang, but controlled development. Integrated existing PHP components via gRPC, used Kubernetes on AWS for orchestration, and CI/CD via GitHub and GitLab.

Abstracted Stripe, Braintree, Adyen, Paymentwall, Chargebee, and Amazon Pay through clean API layers, making each provider swappable without affecting the overall system. Grafana, Prometheus, CloudWatch, and Sentry for observability from the start.

New providers can be integrated in days instead of weeks. No more "fear deployments".