Benchmark max concurrent players

[pierd]

It would be nice to have a snapshot of performance on the website or in the book so that everyone knows what can be achieved with Ambient.

I only have that very rough smoke test I did months ago and it didn't even test a real game - it was just to double check if we can handle ~100 concurrent connections. We could start with:

1. a baseline - an empty project (one created with ambient new).
2. something that at least has input messages (not only ECS diffs) and maybe some processing on the server (a physics simulation running) - based on third party controller or some other example so that people know what to expect in a closer to real game scenario.

For all cases we want to answer the question how many concurrent connection a single Ambient server can sustain while still running the simulation reliably at 60 FPS.

[ten3roberts]

I know @droqen has been working on these benchmarks, specifically for the up and coming wasm runtime improvements

[pierd]

I know @droqen has been working on these benchmarks, specifically for the up and coming wasm runtime improvements

I believe these are different. This issue is more about the size of a game in terms of max concurrent players while @droqen 's benchmarks stress test the client.

Updated the title for clarity.

[droqen]

After a conversation w @pierd I have an idea for how to tackle this

Basic requirement for a benchmark that would be useful for benchmarking max concurrent players

• Simulate clientside inputs, so that any # can be run

My idea

• Boids where each boid is an individual automated client

[itsarjunsinh]

A related thing would be to state Ambient's optimal cloud VM type: General purpose vs Memory optimised vs Compute optimised. Including some recommended specs.

[pierd]

That's highly dependent on the game but I agree that having ballpark numbers from the benchmark would be useful.