feat: add binary to benchmark TPS

[SantiagoPittella]

Partially address #1989

THis PR introduces a new binary to run benchmarks against the Miden node: miden-benchmark. It is split in two subcommands:

• create-proofs: which generates proofs and stores them in a file.
• run-benchmark: that reads the proofs file, submits them to the node and generates a report of the result.

Along with the new binary, this PR also introduces a small change in the block producer configurations by batch_workers configurable.

Note

Currently it won't work against the deployed against the deployed networks due to a change in the protobuf definition of SyncMmr.

Note

Generating 2k transaction (1k mint 1k consume) takes a bit over an hour in a Macbook M4 Max.

In the README of the new binary I added a section about how to start up the block producer with some parameters changes to increase the limit of TPS imposed by the node: TX PER BATCH * BATCHES PER BLOCK / INTERVAL PER BLOCK. At the moment of posting this and with the node's configuration from the README, I didn't reached the limit of TPS yet, though I never sent more than 1k transactions at a time (because the mint transaction are submitted sequentially).

The next steps are to:

• Increase the pool of transaction types used;
• Reach the max TPS limit;
• Try this with testnet's database.

Changelog

[[entry]]
scope       = "node"
impact      = "added"
description = "Added `miden-benchmark` binary for end-to-end TPS measurements. `create-proofs` generates locally-proven mint/consume transaction pairs bound to the target node's chain tip; `run-benchmark` submits the bundle and reports peak/mean/window-average TPS plus inclusion latency, all derived from block-header data."

[[entry]]
scope       = "node"
impact      = "added"
description = "Added `--batch.workers` flag (env `MIDEN_NODE_BLOCK_PRODUCER_BATCH_WORKERS`) to the block-producer to make the batch-builder worker pool size configurable; default remains 2."

[SantiagoPittella]

@bobbinth @Mirko-von-Leipzig this isn't fully working with the deployed networks and remote prover yet because of the version mismatch of the protocol. Do you prefer that I rebase the changes to main? Or should I wait for the new release to land before testing against that? At the moment this is working fine on local networks

[bobbinth]

I think should be fine to keep as is as we'll try to make a release (at least to devnet) towards the end of this week.

[JereSalo]

I ran the benchmarks with the bench-local.sh script and it worked fine 😄

[JereSalo]

Suggested change

	`--batch.workers` (env `MIDEN_NODE_BLOCK_PRODUCER_BATCH_WORKERS`) sets how many batches the block-producer keeps proving
	`--batch.workers` (env `MIDEN_NODE_BATCH_WORKERS`) sets how many batches the block-producer keeps proving

node/bin/node/src/commands/block_producer.rs

Line 151 in 330e709

env = "MIDEN_NODE_BATCH_WORKERS",

[JereSalo]

Do we want to publish this to crates.io or is it meant to be internal tooling? I don't know if it was intentional but in case it wasn't we can set publish to false.

Suggested change

	publish = true
	publish = false

[JereSalo]

Would bumping this number here widen coverage?

Suggested change

	N_TXS: "1"
	N_TXS: "5"

[JereSalo]

It'd be good to add to the .gitignore this directory because when I ran the script (scripts/bench-local.sh) it was created and left untracked.

Suggested change

	/bench-local-run

[JereSalo]

I think part of this "Starting the node" section is stale. For example, the make compose-* targets don't exist anymore, and the miden-node store start / block-producer start / rpc start commands are gone as well.