perf(connector): add integration benchmark for nexmark parsing

[BugenZhao]

Signed-off-by: Bugen Zhao [email protected]I hereby agree to the terms of the RisingWave Labs, Inc. Contributor License Agreement.

What's changed and what's your intention?

To help us investigate #12959.

The "integration" means that we are not only evaluating the performance of parser itself (which is already covered by #9195), but also include the related procedure like tracing, offset maintaining, chunk building and so on, which is just like how we do parsing in production.

---

Something that made me confused (UPDATE: Resolved but still confused)

risingwave/src/connector/benches/nexmark_integration.rs

Lines 117 to 122 in 1ca0a80

	// Enable tracing globally.
	//
	// TODO: we should use `tracing::with_default` to set the dispatch in the scope,
	// so that we can compare the performance with/without tracing side by side.
	// However, why the global dispatch performs much worse than the scoped one?
	dispatch.init();

---

Results:

parse_nexmark           time:   [1.2333 ms 1.2371 ms 1.2409 ms]

parse_nexmark_with_tracing
                        time:   [2.0007 ms 2.0038 ms 2.0071 ms]

...which confirms #12959 (comment).

Checklist

• I have written necessary rustdoc comments
• I have added necessary unit tests and integration tests
• All checks passed in ./risedev check (or alias, ./risedev c)

Documentation

• My PR needs documentation updates. (Please use the Release note section below to summarize the impact on users)

Release note

If this PR includes changes that directly affect users or other significant modifications relevant to the community, kindly draft a release note to provide a concise summary of these changes. Please prioritize highlighting the impact these changes will have on users.

[xxchan]

Should it be like this?

Suggested change

	// Enable tracing globally.
	//
	// TODO: we should use `tracing::with_default` to set the dispatch in the scope,
	// so that we can compare the performance with/without tracing side by side.
	// However, why the global dispatch performs much worse than the scoped one?
	dispatch.init();
	c.bench_function("parse_nexmark", |b| {
	b.iter_batched(
	make_stream_iter,
	|mut iter| iter.next().unwrap(),
	BatchSize::SmallInput,
	)
	});
	tracing::dispatcher::with_default(&dispatch, || {
	c.bench_function("parse_nexmark_with_tracing_scoped", |b| {
	b.iter_batched(
	make_stream_iter,
	|mut iter| iter.next().unwrap(),
	BatchSize::SmallInput,
	)
	})
	});

[BugenZhao]

If we only have single bench function of parse_with_tracing_scoped, then it's enabled correctly. 😄 If two are compiled together, then it seems to be disabled.

[BugenZhao]

Resolved in 456cd32, but not sure why. 😄

[codecov]

Codecov Report

Merging #13073 (456cd32) into main (efdf3c9) will decrease coverage by 0.07%.
Report is 4 commits behind head on main.
The diff coverage is n/a.

@@            Coverage Diff             @@
##             main   #13073      +/-   ##
==========================================
- Coverage   68.40%   68.33%   -0.07%     
==========================================
  Files        1498     1498              
  Lines      252154   252154              
==========================================
- Hits       172490   172320     -170     
- Misses      79664    79834     +170     

Flag	Coverage Δ
rust	68.33% <ø> (-0.07%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

see 21 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more