Solaris performs massive amounts of processes, including asynchronous tasks that require notification upon completion. They employ a publisher-subscriber model to deliver millions of messages per day to partners, but their legacy solution faced limitations in scalability. To address this, they redesigned their architecture to prioritize robustness, fast delivery, and containment of retries, utilizing serverless components like SNS, SQS, Lambda, and DynamoDB. They also conducted performance tests and learned important lessons about error handling, batch processing, and parallelization. As a result, their system now successfully handles millions of messages per day with impressive latency metrics.
















