This summarizes how AWS powered Amazon Prime Day 2026 and the concrete infrastructure metrics that kept the shopping event available, fast, and secure. Multiple AWS services handled unprecedented traffic: compute ran heavily on EC2 with Graviton supplying up to 49% of Amazon.com’s EC2 capacity, serverless Lambda handled about 2.3 trillion invocations per day, and container workloads averaged 158.3 million ECS tasks per day on Fargate (a 47.7% year-over-year rise). Content and delivery scaled globally via CloudFront (over 2.1 trillion HTTP requests for the week), while databases and caches sustained extreme load - DynamoDB processed roughly 59 trillion requests across June 23-26 with peaks of 192 million req/s, Aurora ran hundreds of billions of transactions storing ~5,491 TB, and ElastiCache served trillions to quadrillions of requests.
Observability, streaming, messaging, and resilience also scaled dramatically: CloudWatch handled about 2.15 quadrillion metric observations per day, CloudTrail logged 3.6 trillion API events across four days (up 44% vs. the prior year), GuardDuty monitored ~14.08 trillion log events per hour (up 59%), Kinesis peaked at 988 million records/sec, SNS delivered ~5 trillion messages in a single day, and SQS peaked at 213 million messages/sec. Operational readiness included over 44,000 Fault Injection Service experiments (6x 2025), and AWS offered Countdown Premium to help other organizations prepare for similar high‑traffic events.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.