chore(deps): add OTLP ingest throughput benchmark and tokio regression repro

[thieman]

Summary

Adds benchmark infrastructure to reproduce and measure the ~25% throughput regression introduced by tokio 1.51 (PR #7431 — "steal tasks from the LIFO slot").

Running the benchmark

make bench-tokio-regression

This builds ADP and millstone in release mode, runs 3 warm-up + 10 measured rounds of OTLP trace ingest via gRPC against tokio 1.50 and 1.51, and prints a summary:

┌─────────────────────────────────────────────────────────┐
│  tokio LIFO-stealing throughput regression (PR #7431)   │
│  OTLP trace ingest via gRPC — 10 runs each, 3 warmup    │
└─────────────────────────────────────────────────────────┘

▶ tokio 1.50.0  (building + running 3 warmup + 10 measured...)
  n=10  mean=3.0 MB/s  median=3.0 MB/s  stdev=0.1 MB/s  min=2.9  max=3.3

▶ tokio 1.51.0  (building + running 3 warmup + 10 measured...)
  n=10  mean=2.2 MB/s  median=2.3 MB/s  stdev=0.1 MB/s  min=2.0  max=2.4

┌─────────────────────────────────────┐
│  tokio 1.50.0    median:    3.0 MB/s │
│  tokio 1.51.0    median:    2.3 MB/s │
│                                     │
│  regression:                  -23%  │
└─────────────────────────────────────┘

The regression is most visible on Linux with 4+ cores (~25%). On macOS the effect is smaller (~10%) due to better cache coherency.

What's included

• test/bench/bench-tokio-regression.sh — regression repro driver
• test/bench/otlp-ingest.sh — single-version benchmark script (also usable standalone)
• test/bench/adp-otlp.yaml — minimal ADP config (standalone mode, OTLP enabled)
• test/bench/millstone-otlp.yaml — load corpus matching the SMP otlp_ingest_traces test
• test/bench/intake-blackhole.py — fake HTTP intake to absorb forwarded traces
• lib/saluki-components/benches/otlp_traces.rs — Criterion benchmarks for decode, translate, async pipeline, and gRPC ingest hot paths
• lib/saluki-components/benches/scheduler.rs — scheduler sensitivity benchmark targeting LIFO slot behavior

Requirements

Linux with 4+ cores, Rust toolchain, Python 3, openssl, nc, fuser.

🤖 Generated with Claude Code

[pr-commenter]

Binary Size Analysis (Agent Data Plane)

Target: c4787c3 (baseline) vs 4a72828 (comparison) diff
Analysis Type: Stripped binaries (debug symbols excluded)
Baseline Size: 37.09 MiB
Comparison Size: 37.00 MiB
Size Change: -97.11 KiB (-0.26%)
Pass/Fail Threshold: +5%
Result: PASSED ✅

Changes by Module

Module	File Size	Symbols
figment	-112.72 KiB	39
core	+14.37 KiB	361
piecemeal	+13.19 KiB	18
saluki_common::task::instrument	-9.54 KiB	19
http_body_util	-8.89 KiB	12
datadog_protos::trace_piecemeal_include::datadog	-8.47 KiB	13
[sections]	-8.41 KiB	5
memory_accounting::allocator::Tracked	+7.08 KiB	19
saluki_components::common::datadog	+5.82 KiB	17
tower	-4.74 KiB	4
http	+4.37 KiB	19
bytes	+4.33 KiB	9
saluki_io::net::util	+3.81 KiB	4
tokio	+2.81 KiB	43
otlp_protos::otlp_include::opentelemetry	-2.77 KiB	19
saluki_components::sources::dogstatsd	+2.50 KiB	8
prost	+2.42 KiB	17
anon.c558e9f1bd059043dd8639ac17522c07.403.llvm.13126592839154015127	-1.93 KiB	1
anon.b39606bfe243aa45c40b54889fc520b2.2.llvm.14371567054272567064	+1.93 KiB	1
[Unmapped]	+1.92 KiB	1

Detailed Symbol Changes

    FILE SIZE        VM SIZE    
 --------------  -------------- 
  [NEW] +12.3Ki  [NEW] +12.1Ki    _<core::marker::PhantomData<T> as serde_core::de::DeserializeSeed>::deserialize::hf32244927aa7c630
  [NEW] +9.19Ki  [NEW]    +355    _<figment::value::de::MapDe<D,F> as serde_core::de::MapAccess>::next_value_seed::hbb74e34b2ac40b91
  [NEW] +6.50Ki  [NEW] +6.28Ki    saluki_components::common::datadog::apm::_::_<impl serde_core::de::Deserialize for saluki_components::common::datadog::apm::ApmConfiguration>::deserialize::h13cd270478ab72a7
  [NEW] +4.29Ki  [NEW] +4.07Ki    _<http::header::map::HeaderMap<T> as core::iter::traits::collect::Extend<$LP$core::option::Option<http::header::name::HeaderName>,T$RP$>>::extend::he53653d51168fbc4
  [DEL] -4.34Ki  [DEL] -4.18Ki    _<figment::value::de::ConfiguredValueDe<I> as serde_core::de::Deserializer>::deserialize_struct::hc8570a6db8448a31
  [DEL] -4.36Ki  [DEL] -4.21Ki    _<figment::value::magic::Tagged<T> as figment::value::magic::Magic>::deserialize_from::h3aafd8a1f2fd337b
 -41.6% -4.38Ki -41.9% -4.38Ki    h2::codec::framed_write::Encoder<B>::buffer::hd1cc24eaea5bb719
  [DEL] -4.54Ki  [DEL] -4.38Ki    _<figment::value::de::ConfiguredValueDe<I> as serde_core::de::Deserializer>::deserialize_struct::h12e2df1fb5d3b706
 -92.2% -4.69Ki -95.0% -4.69Ki    datadog_protos::trace_piecemeal_include::datadog::trace::AgentPayloadBuilder<S>::add_tracer_payloads::h452beefbd5c34217
  [DEL] -4.80Ki  [DEL] -4.65Ki    _<figment::value::magic::Tagged<T> as figment::value::magic::Magic>::deserialize_from::h60edcbe0028cd4b1
  [DEL] -5.08Ki  [DEL] -4.96Ki    figment::value::de::_<impl figment::value::value::Value>::deserialize_from::h2987322268e98bdb
  [DEL] -5.27Ki  [DEL] -5.12Ki    _<figment::value::magic::RelativePathBuf as figment::value::magic::Magic>::deserialize_from::h7bb132a2aab4941b
  [DEL] -5.86Ki  [DEL] -5.71Ki    _<figment::value::magic::RelativePathBuf as figment::value::magic::Magic>::deserialize_from::h35c2ddc84d0593f2
  [DEL] -6.26Ki  [DEL] -6.11Ki    _<figment::value::magic::Tagged<T> as figment::value::magic::Magic>::deserialize_from::h05666e9a1e35fbff
  [DEL] -6.76Ki  [DEL] -6.61Ki    _<figment::value::magic::RelativePathBuf as figment::value::magic::Magic>::deserialize_from::h3ee77dc250716eaa
  [DEL] -7.73Ki  [DEL] -7.61Ki    figment::value::de::_<impl figment::value::value::Value>::deserialize_from::h71c55d9a8e477b5c
  [DEL] -11.4Ki  [DEL]    -355    _<figment::value::de::MapDe<D,F> as serde_core::de::MapAccess>::next_value_seed::h4c05907830883f1b
  [DEL] -11.7Ki  [DEL] -11.5Ki    _<figment::value::de::ConfiguredValueDe<I> as serde_core::de::Deserializer>::deserialize_struct::h513f5e56d7a1cbaf
  -0.3% -11.7Ki  -0.3% -8.13Ki    [3241 Others]
  [DEL] -14.9Ki  [DEL] -14.7Ki    _<figment::value::magic::Tagged<T> as figment::value::magic::Magic>::deserialize_from::h1e3e1e520bcca7e7
  [DEL] -15.6Ki  [DEL] -15.5Ki    _<figment::value::magic::RelativePathBuf as figment::value::magic::Magic>::deserialize_from::h39b38e9ac37ee08e
  -0.3% -97.1Ki  -0.3% -89.9Ki    TOTAL

[pr-commenter]

Regression Detector (Agent Data Plane)

Regression Detector Results

Run ID: dabc6137-95f1-428b-b2ba-9a1e3979fe23

Baseline: c4787c3
Comparison: 4a72828
Diff

Optimization Goals: ✅ No significant changes detected

Experiments ignored for regressions

Regressions in experiments with settings containing erratic: true are ignored.

perf	experiment	goal	Δ mean %	Δ mean % CI	trials	links
➖	otlp_ingest_logs_5mb_cpu	% cpu utilization	+2.96	[-1.96, +7.88]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_logs_5mb_throughput	ingress throughput	-0.02	[-0.15, +0.11]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_logs_5mb_memory	memory utilization	-4.25	[-4.54, -3.97]	1	(metrics) (profiles) (logs)

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI	trials	links
➖	otlp_ingest_logs_5mb_cpu	% cpu utilization	+2.96	[-1.96, +7.88]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_metrics_5mb_cpu	% cpu utilization	+1.73	[-5.66, +9.12]	1	(metrics) (profiles) (logs)
➖	dsd_uds_500mb_3k_contexts_throughput	ingress throughput	+1.11	[+0.98, +1.24]	1	(metrics) (profiles) (logs)
➖	dsd_uds_512kb_3k_contexts_memory	memory utilization	+0.44	[+0.30, +0.58]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_traces_5mb_memory	memory utilization	+0.24	[+0.07, +0.40]	1	(metrics) (profiles) (logs)
➖	dsd_uds_500mb_3k_contexts_memory	memory utilization	+0.21	[+0.07, +0.36]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_traces_ottl_transform_5mb_memory	memory utilization	+0.20	[+0.04, +0.36]	1	(metrics) (profiles) (logs)
➖	dsd_uds_1mb_3k_contexts_memory	memory utilization	+0.17	[+0.03, +0.32]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_traces_ottl_transform_5mb_throughput	ingress throughput	+0.17	[+0.10, +0.24]	1	(metrics) (profiles) (logs)
➖	dsd_uds_500mb_3k_contexts_cpu	% cpu utilization	+0.15	[-1.38, +1.68]	1	(metrics) (profiles) (logs)
➖	quality_gates_rss_dsd_ultraheavy	memory utilization	+0.11	[-0.02, +0.24]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_traces_ottl_filtering_5mb_throughput	ingress throughput	+0.08	[+0.00, +0.15]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_traces_ottl_filtering_5mb_memory	memory utilization	+0.06	[-0.19, +0.31]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_metrics_5mb_throughput	ingress throughput	+0.03	[-0.14, +0.19]	1	(metrics) (profiles) (logs)
➖	dsd_uds_100mb_3k_contexts_throughput	ingress throughput	+0.00	[-0.04, +0.05]	1	(metrics) (profiles) (logs)
➖	dsd_uds_1mb_3k_contexts_throughput	ingress throughput	+0.00	[-0.06, +0.06]	1	(metrics) (profiles) (logs)
➖	dsd_uds_512kb_3k_contexts_throughput	ingress throughput	-0.00	[-0.06, +0.05]	1	(metrics) (profiles) (logs)
➖	dsd_uds_10mb_3k_contexts_throughput	ingress throughput	-0.01	[-0.16, +0.15]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_logs_5mb_throughput	ingress throughput	-0.02	[-0.15, +0.11]	1	(metrics) (profiles) (logs)
➖	quality_gates_rss_dsd_medium	memory utilization	-0.11	[-0.28, +0.06]	1	(metrics) (profiles) (logs)
➖	quality_gates_rss_idle	memory utilization	-0.12	[-0.17, -0.06]	1	(metrics) (profiles) (logs)
➖	dsd_uds_100mb_3k_contexts_memory	memory utilization	-0.15	[-0.30, +0.01]	1	(metrics) (profiles) (logs)
➖	quality_gates_rss_dsd_low	memory utilization	-0.17	[-0.34, +0.00]	1	(metrics) (profiles) (logs)
➖	dsd_uds_10mb_3k_contexts_memory	memory utilization	-0.19	[-0.35, -0.03]	1	(metrics) (profiles) (logs)
➖	quality_gates_rss_dsd_heavy	memory utilization	-0.28	[-0.41, -0.15]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_traces_5mb_throughput	ingress throughput	-0.29	[-0.36, -0.21]	1	(metrics) (profiles) (logs)
➖	dsd_uds_100mb_3k_contexts_cpu	% cpu utilization	-0.46	[-6.35, +5.44]	1	(metrics) (profiles) (logs)
➖	dsd_uds_10mb_3k_contexts_cpu	% cpu utilization	-0.48	[-31.34, +30.39]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_metrics_5mb_memory	memory utilization	-0.50	[-0.63, -0.37]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_traces_ottl_transform_5mb_cpu	% cpu utilization	-0.91	[-2.97, +1.14]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_traces_ottl_filtering_5mb_cpu	% cpu utilization	-1.35	[-3.55, +0.85]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_traces_5mb_cpu	% cpu utilization	-2.30	[-4.25, -0.35]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_logs_5mb_memory	memory utilization	-4.25	[-4.54, -3.97]	1	(metrics) (profiles) (logs)
➖	dsd_uds_512kb_3k_contexts_cpu	% cpu utilization	-4.70	[-61.01, +51.61]	1	(metrics) (profiles) (logs)
➖	dsd_uds_1mb_3k_contexts_cpu	% cpu utilization	-8.71	[-60.13, +42.70]	1	(metrics) (profiles) (logs)

Bounds Checks: ✅ Passed

perf	experiment	bounds_check_name	replicates_passed	observed_value	links
✅	quality_gates_rss_dsd_heavy	memory_usage	10/10	119.45MiB ≤ 140MiB	(metrics) (profiles) (logs)
✅	quality_gates_rss_dsd_low	memory_usage	10/10	39.75MiB ≤ 50MiB	(metrics) (profiles) (logs)
✅	quality_gates_rss_dsd_medium	memory_usage	10/10	61.74MiB ≤ 75MiB	(metrics) (profiles) (logs)
✅	quality_gates_rss_dsd_ultraheavy	memory_usage	10/10	177.01MiB ≤ 200MiB	(metrics) (profiles) (logs)
✅	quality_gates_rss_idle	memory_usage	10/10	27.15MiB ≤ 40MiB	(metrics) (profiles) (logs)

Explanation

Confidence level: 90.00%
Effect size tolerance: |Δ mean %| ≥ 5.00%

Performance changes are noted in the perf column of each table:

• ✅ = significantly better comparison variant performance
• ❌ = significantly worse comparison variant performance
• ➖ = no significant change in performance

A regression test is an A/B test of target performance in a repeatable rig, where "performance" is measured as "comparison variant minus baseline variant" for an optimization goal (e.g., ingress throughput). Due to intrinsic variability in measuring that goal, we can only estimate its mean value for each experiment; we report uncertainty in that value as a 90.00% confidence interval denoted "Δ mean % CI".

For each experiment, we decide whether a change in performance is a "regression" -- a change worth investigating further -- if all of the following criteria are true:

1. Its estimated |Δ mean %| ≥ 5.00%, indicating the change is big enough to merit a closer look.

2. Its 90.00% confidence interval "Δ mean % CI" does not contain zero, indicating that if our statistical model is accurate, there is at least a 90.00% chance there is a difference in performance between baseline and comparison variants.

3. Its configuration does not mark it "erratic".