Generation of Leios node resource usage data from validated Haskell prototype

[Saizan]

Goals

1. Run simulations for different orders of magnitude of throughput scenarios, providing network and CPU usage events and/or metrics useful for calculating running node costs. Compare data to the high-level cost estimator tool.

2. Demonstrate adherence to the Short Leios protocol, also with the help of the formal specification.

3. Within reason, incorporate new protocol variations from research workstream.

4. If time permits, document mini-protocols design and solicit review from the Cardano Networking team.

5. Build confidence in simulation results by comparison to an idealized diffusion model: we would see how close the simulation gets by disabling features the model ignores, and then measure the impact of re-enabling each feature.

Definition of done

Numbered to match the goals:

1. Simulation trace contains events for the CPU usages of interest (cryptography operations and block validation) as specified in configuration, and also events for messages exchanged between nodes.
2. Simulation trace contains Node activity events that can be compared to the state transitions of the formal specification, and conformance can be demonstrated through analysis and/or testing.
3. Ongoing task.
4. Draft of technical report comparable to mini-protocols section of the report for the current cardano-node.
5. Written report describing the idealized diffusion model and the sources of discrepancy with the simulation.

[njd42]

This sounds like a very sensible and valuable approach. Creating idealised models helps with general socialisation as well as establishing reasonable 'stretch goals' during development.

Sharing the mini-protocols with the network team would be a good source of peer review

[will-break-it]

I see all of the points as quite valuable but I am a bit concerned about the overall scope and whether the number of goals are achievable given the timeframe.

3. can vary a lot and will highly depend on progress made on short Leios as well as how the rest of the team progresses. If we start to build full Leios features, I'd like to see other workstreams do so as well.
4. I don't exactly understand how much work this is. From previous discussions it sounded like the newly introduced block fetch protocols are very similar to what Praos is already using. Maybe we can discuss a bit more about it.

I see 1., 2. and 5. currently as most valuable in the short term but I am open to reprioritize.

cc @bwbush