AWS Lambda Web Functions: what changes when you deploy an HTTP application

On October 12, AWS announced a Lambda mode for running conventional HTTP servers, including applications built with Express, Fastify, or Next.js.

Three infrastructure details deserve attention:

• Concurrency: one environment can handle up to 128 simultaneous requests.
• Active CPU: CPU billing reflects the processing actually used. Memory is billed for as long as the environment exists.
• Distribution: you can deploy code to multiple regions and configure latency-based routing (see the source).

To evaluate a migration, I’d measure:

1. p95 latency under load.
2. Simultaneous connections to the database.
3. Distance between compute and data.
4. Total cost under a representative traffic pattern.

Being able to share an environment across requests also means reviewing how we handle in-memory state and shared resources.

What would you measure first to decide whether this model fits your application?