perf(encode): make the encoding bench reflect a real Node.js HTTP request#8668
Merged
BridgeAR merged 2 commits intoMay 28, 2026
Merged
Conversation
…quest The previous fixture was 30 copies of one span with three meta entries and three metrics, so span 1 warmed the string cache and every later iteration replayed identical state -- anything that moved per-span work beyond cache filling read as noise. The new fixture is a 30-span Express-request shape (server root + middleware fan + Postgres / Redis / HTTP-client / DNS spans) with one error span carrying a 1.5 KB `error.stack`. Hot keys (`span.kind`, `component`, `runtime-id`, env, version) reuse the same strings across spans so the cache still warms after one iteration but the meta-map hot loop and the wire output run on production-shaped data.
🎉 All green!🧪 All tests passed 🎯 Code Coverage (details) 🔗 Commit SHA: a5cd63b | Docs | Datadog PR Page | Give us feedback! |
Contributor
Overall package sizeSelf size: 5.96 MB Dependency sizes| name | version | self size | total size | |------|---------|-----------|------------| | import-in-the-middle | 3.0.1 | 82.56 kB | 817.39 kB | | opentracing | 0.14.7 | 194.81 kB | 194.81 kB | | dc-polyfill | 0.1.11 | 25.74 kB | 25.74 kB |🤖 This report was automatically generated by heaviest-objects-in-the-universe |
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## master #8668 +/- ##
==========================================
- Coverage 40.95% 39.78% -1.17%
==========================================
Files 286 495 +209
Lines 15987 30639 +14652
Branches 3532 3532
==========================================
+ Hits 6548 12191 +5643
- Misses 9439 18448 +9009 Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
BridgeAR
changed the title
BenchmarksBenchmark execution time: 2026-05-27 21:13:28 Comparing candidate commit a5cd63b in PR branch Found 0 performance improvements and 58 performance regressions! Performance is the same for 1434 metrics, 101 unstable metrics. scenario:encoders-0.4-20
scenario:encoders-0.4-22
scenario:encoders-0.4-24
scenario:encoders-0.4-events-legacy-20
scenario:encoders-0.4-events-legacy-22
scenario:encoders-0.4-events-legacy-24
scenario:encoders-0.4-events-native-20
scenario:encoders-0.4-events-native-22
scenario:encoders-0.4-events-native-24
scenario:encoders-0.5-20
scenario:encoders-0.5-22
scenario:encoders-0.5-24
scenario:encoders-0.5-events-legacy-20
scenario:encoders-0.5-events-legacy-22
scenario:encoders-0.5-events-legacy-24
|
The previous fixture's trace was built once and reused for 5000 encodes unchanged: the encoder's string cache, magnitude-branch predictor, and per-span ID reads all settled to a single state and held it, so any optimisation that depended on hot-vs-cold variance read as noise. `tickTrace(trace, iteration)` now runs before each encode and rewrites every per-request dynamic field in place (timestamps, durations, ID buffer low halves, `db.row_count`, root-span route / URL / resource / status / client IP, and error type / message / stack). Hot strings reused across spans stay constant so the cache still warms after one iteration. The `+ iteration * 4096` step on `span.start` is the smallest increment visible past the IEEE-754 double's ULP at the ~1.7e18 nano-timestamp magnitude (256 nanos). A real fix for that precision loss needs the span to carry `start` as `BigInt`, which is a separate change on the tracer side, not here.
pabloerhard
approved these changes
May 28, 2026
dd-octo-sts Bot
pushed a commit
that referenced
this pull request
May 29, 2026
…quest (#8668) * bench(encode): make the encoding bench reflect a real Node.js HTTP request The previous fixture was 30 copies of one span with three meta entries and three metrics, so span 1 warmed the string cache and every later iteration replayed identical state -- anything that moved per-span work beyond cache filling read as noise. The new fixture is a 30-span Express-request shape (server root + middleware fan + Postgres / Redis / HTTP-client / DNS spans) with one error span carrying a 1.5 KB `error.stack`. Hot keys (`span.kind`, `component`, `runtime-id`, env, version) reuse the same strings across spans so the cache still warms after one iteration but the meta-map hot loop and the wire output run on production-shaped data. * bench(encode): vary every per-request field across iterations The previous fixture's trace was built once and reused for 5000 encodes unchanged: the encoder's string cache, magnitude-branch predictor, and per-span ID reads all settled to a single state and held it, so any optimisation that depended on hot-vs-cold variance read as noise. `tickTrace(trace, iteration)` now runs before each encode and rewrites every per-request dynamic field in place (timestamps, durations, ID buffer low halves, `db.row_count`, root-span route / URL / resource / status / client IP, and error type / message / stack). Hot strings reused across spans stay constant so the cache still warms after one iteration. The `+ iteration * 4096` step on `span.start` is the smallest increment visible past the IEEE-754 double's ULP at the ~1.7e18 nano-timestamp magnitude (256 nanos). A real fix for that precision loss needs the span to carry `start` as `BigInt`, which is a separate change on the tracer side, not here.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
The old encoding bench was 30 clones of one span with three meta keys
(
a/hello/and) and three metrics. Span 1 filled the string cache;every iteration after that replayed identical state, so anything that
moved per-span work beyond cache filling read as noise.
The new fixture (
trace-fixture.js) is a 30-span Express-request shape:server root, middleware fan, a handful of Postgres / Redis / HTTP-client
spans, two DNS lookups, and one error span carrying a 1.5 KB
error.stack. Hot keys (span.kind,component,runtime-id, env,version) reuse the same strings across spans, so the cache still warms
after one iteration but the meta-map hot loop and the wire output run
on production-shaped data.
Why
The sirun bench-comparison comment reads this fixture as a baseline
regression in any PR that introduces it -- the new fixture does roughly
5x the encode work per iteration. Landing it on its own absorbs the
step once, so encoder-side PRs that follow can be compared apples to
apples instead of new-fixture-vs-old-fixture.