Mercure Hub Load Testing - Clarification on 40k Concurrent Connections Benchmark

[rishabakalankan7]

I'm setting up a single-instance Mercure hub and performing load testing using the gatling test suite provided by the mercure team. The official documentation states that a single t3.micro instance can handle approximately 40,000 concurrent connections.

I'd like to understand the specifics of this benchmark:

1. Does the 40k figure refer to subscribers only, or does it include both subscribers and publishers?
2. Which transport was used during this benchmark (local, Bolt, Redis, etc.)?
3. Was TLS enabled during this load testing?

Any insights or links to the original benchmarking would be appreciated.

Environment:

• Mercure Hub (single instance) in AWS EC2(t3.small)