Can we use AWS Lambda to handle high-concurrency event processing without hitting scaling limits?

All answers to this question.

Handling 50k RPS in Cloud Computing (AWS/Azure/GCP) requires more than just "going serverless." AWS Lambda has a default burst limit that varies by region. For your scale, Provisioned Concurrency is mandatory to avoid "Cold Starts" during the peak of the sale. However, you should also implement SQS as a buffer. By putting a queue in front of your Lambda, you can smooth out the spikes and process the events at a consistent rate without hitting the account-level concurrency ceiling. Fargate is great, but Lambda is much easier to manage if your logic is short-lived.

   Answered 2025-02-13 by Sandra Jenkins

Are you concerned about the downstream database? Even if Lambda scales to 50k, can your RDS instance handle that many concurrent connections without falling over?

   Answered 2025-02-15 by Paul Roberts

• You're spot on, Paul. We are using RDS Proxy to manage connection pooling. Without it, the database would definitely crash within seconds. We are also looking at DynamoDB for the core transaction data because it handles massive horizontal scaling much more natively than a relational database, which is a key consideration when you're designing for such high-intensity short-duration traffic bursts.

     Commented 2025-02-16 by Edward King

Check your account limits early. AWS can increase your concurrency limits if you provide a valid business case and a few weeks' notice before the event.

   Answered 2025-02-17 by Lisa Turner

• Exactly, Lisa. Proactive communication with AWS support is the best way to ensure your infrastructure doesn't throttle right when the sale goes live.

     Commented 2025-02-18 by Larry Watkins