feat(bench): benchmark native MPP opens and settlements

[decofe]

Replace the deployed-contract approve/open/close workload in mpp.yml with a native TIP-20 Channel Reserve open -> signed settlement sequence.

• Bind a payer and payee per channel, save the signed open, and use its actual descriptor and transaction context for the settlement voucher.
• Pay the full deposit to the payee in the second transaction. No deployment or token approval is needed; the settled channel state remains present.
• Preserve inclusion ordering between the two steps. The preset is 50% opens and 50% settlements by transaction count; mix.yml is unchanged.

Dependency

Requires txgen #200, which adds sequence-step save and native mpp_settle support. Merge and publish that change before using this preset with the default txgen image. For a pre-merge bench run, use txgen-ref=57a891b99bc8a07b44bc0a67b2d98000c256e00c.

Validation

Resolved the preset through the benchmark helper, then generated and executed 100 transactions across three accounts on a fresh local Tempo v1.14.0 dev node:

• 50 distinct native channel opens and 50 successful settlements; zero reverted receipts.
• Every open included in an earlier block than its settlement.
• Every channel ended with settled = deposit = 1 and no close request.
• Verified the settlement events and token transfers to each payee, zero reserve allowances, and zero final reserve token balance.

This is a local correctness smoke test, not a throughput comparison.

Prompted by: @shekhirin

[tempo-github-sts]

cc @decofe

⚪ Benchmark complete: No Difference View job

⚪ Bench Comparison: No Difference

Refs: 07761a7 vs 96360a4
Criteria: 95% run-bootstrap CI must clear floor; cells show delta (+/-CI/floor).

Configuration

• Derek command: derek bench mode=e2e preset=mpp duration=30 bloat=1 token-count=4 tps=1000 accounts=100 max-concurrent-requests=100 baseline=07761a78a4ac00988533aa8acbcb6667786b625d feature=96360a4b13c62eec65f75a0978cf9b6f0ca72072 baseline-hardfork="" feature-hardfork="" gas-limit=1000000000000 run-pairs=1 run-side=comparison otlp=false metrics=false no-cache=false force-bloat=false general-gas-limit=1000000000000 txgen-ref=57a891b99bc8a07b44bc0a67b2d98000c256e00c
• Bloat: 1000 MiB
• Token count: 4
• Preset: mpp
• Target TPS: 1000
• Duration: 30s
• Run pairs: 1
• Baseline blocks: 60
• Feature blocks: 59

Tempo Metrics

Metric	Baseline	Feature	Delta
TPS Mean	782	788	+0.77% ⚪
Gas Throughput [Mgas/s]	149.3	150.5	+0.80% ⚪
Block Time Mean [ms]	501.6	502.8	+0.24% ⚪
Block Time P50 [ms]	502.0	502.0	+0.00% ⚪
Block Time P90 [ms]	506.0	508.0	+0.40% ⚪
Block Time P99 [ms]	508.0	511.0	+0.59% ⚪

Builder

Metric	Baseline	Feature	Delta
Gas Throughput [Mgas/s]	159.4	160.1	+0.41% ⚪
P50 [ms]	480.1	481.9	+0.37% ⚪
P90 [ms]	481.8	483.0	+0.25% ⚪
P99 [ms]	483.9	483.4	-0.10% ⚪

Builder details

Metric	Baseline	Feature	Delta
Finish P50 [ms]	0.3	0.3	+0.00%
Finish P90 [ms]	0.4	0.4	+0.00%
Finish P99 [ms]	2.6	0.8	-69.23%
Pool Fetch P50 [ms]	0.0	0.0	0.00%
Pool Fetch P90 [ms]	0.1	0.1	+0.00%
Pool Fetch P99 [ms]	0.3	0.9	+200.00%
Stop Reason — Build Budget	57	56	-1.75%
Serialized Block Size P50 [KiB]	160.9	228.5	+41.97%
Serialized Block Size P90 [KiB]	242.0	232.0	-4.12%
Serialized Block Size P99 [KiB]	246.0	233.2	-5.19%
Serialized Block Size / Tx P50 [B/tx]	497.2	496.7	-0.10%
Serialized Block Size / Tx P90 [B/tx]	498.3	497.9	-0.08%
Serialized Block Size / Tx P99 [B/tx]	586.4	529.4	-9.72%
Fill Idle P50 [ms]	440.0	442.0	+0.45%
Fill Idle P90 [ms]	446.0	446.0	+0.00%
Fill Idle P99 [ms]	447.0	448.0	+0.22%

Validator

Metric	Baseline	Feature	Delta
Gas Throughput [Mgas/s]	5613.3	5677.0	+1.13% ⚪
P50 [ms]	13.4	12.1	-9.70% ⚪
P90 [ms]	17.0	16.4	-3.53% ⚪
P99 [ms]	17.7	18.8	+6.21% ⚪

Observability

• Internal dashboard
• Logs
• Traces

[github-actions]

This PR has been marked stale due to 7 days of inactivity.