Is Serverless PaaS better than traditional managed platforms for high-traffic web applications?

All answers to this question.

For e-commerce, a hybrid "Container-based PaaS" is often the sweet spot. While pure serverless functions (FaaS) scale incredibly well for background tasks, the "cold start" can hurt user experience on your frontend if a visitor hits a cold instance. A traditional PaaS that keeps a minimum number of "warm" instances running provides more consistent latency. In 2024, many platforms offer "Scale-to-Zero" capabilities with "Provisioned Concurrency" to eliminate cold starts, but this can get expensive. If your traffic is consistent but has predictable spikes, a standard PaaS with auto-scaling rules is often cheaper and faster.

   Answered 2024-06-12 by Kimberly White

Does the cost of "Provisioned Concurrency" in serverless end up being higher than just paying for a reserved instance in a traditional PaaS model over a long period?

   Answered 2024-06-14 by Jason Reed

• Usually, yes, Jason. If you have a baseline of traffic that never hits zero, a reserved PaaS instance is significantly more cost-effective. Serverless is only cheaper when your traffic is "bursty" or intermittent. For a major e-commerce brand with constant visitors, paying for the "always-on" nature of a standard PaaS usually saves about 30% compared to high-volume serverless execution with reserved capacity.

     Commented 2024-06-16 by Gregory Evans

We found that managing the cold starts for our Node.js app on serverless was more work than just setting up basic HPA on a managed Kubernetes PaaS.

   Answered 2024-06-17 by Robert Foster

• I agree, Robert. Managed Kubernetes (K8s as a Service) is becoming the modern PaaS of choice because it gives you that serverless feel without the architectural limitations of FaaS.

     Commented 2024-06-20 by Amanda Peterson