perf(sdk-trace-base): use Uint8Array for browser RandomIdGenerator #6209

overbalance · 2025-12-10T23:48:35Z

Summary

Optimizes browser RandomIdGenerator by using Uint8Array with a byte-to-hex lookup table instead of character code generation with String.fromCharCode
Refactors to use proper class methods instead of arrow property assignments
Adds benchmark tests for Node and browser

Benchmark Results

Benchmark	Node	Browser (old)	Browser (new)	Browser Improvement
`generateTraceId`	9.07M ops/sec	2.48M ops/sec	7.17M ops/sec	+189%
`generateSpanId`	13.3M ops/sec	4.70M ops/sec	10.4M ops/sec	+121%
create spans (10 attrs)	1.24M ops/sec	645K ops/sec	903K ops/sec	+40%
BatchSpanProcessor	1.19M ops/sec	694K ops/sec	993K ops/sec	+43%

Test plan

All existing tests pass
Verified output format: 32-char hex for traceId, 16-char for spanId
Benchmarks run on both Node (npm run test:bench) and browser (npm run test:bench:browser)

Benchmark Infrastructure

Separate Node and browser benchmarks

Benchmarks are split into separate files to allow distinct test names for reporting:

test/node/*.bench.ts - Node benchmarks (e.g., generateTraceId)
test/browser/*.bench.ts - Browser benchmarks with "(browser)" suffix (e.g., generateTraceId (browser))

Output files

test/node/.benchmark-results.txt - Node benchmark results
Browser benchmark results are not written to disk, only displayed in the console

Test folder structure

test/node/ - Node-only tests and benchmarks
test/browser/ - Browser-only tests and benchmarks

codecov · 2025-12-10T23:54:07Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.58%. Comparing base (38924cb) to head (3499c8f).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #6209      +/-   ##
==========================================
+ Coverage   95.46%   95.58%   +0.11%     
==========================================
  Files         317      314       -3     
  Lines        9602     9590      -12     
  Branches     2221     2221              
==========================================
  Hits         9167     9167              
+ Misses        435      423      -12

see 3 files with indirect coverage changes

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

BenoitZugmeyer · 2025-12-11T16:46:51Z

packages/opentelemetry-sdk-trace-base/src/platform/browser/RandomIdGenerator.ts

+  for (let i = 0; i < buf.length; i++) {
+    buf[i] = (Math.random() * 256) >>> 0;
+  }
+  // Ensure non-zero
+  for (let i = 0; i < buf.length; i++) {
+    if (buf[i] > 0) return;
+  }
+  buf[buf.length - 1] = 1;


question: did you consider using crypto.getRandomValues?

Without running a test i can't say for sure but I'd say its very unlikely the crypto module is fast enough. The "cryptographically strong" requirement of that module trades off speed and its pretty significant in a lot of cases. We only have need for statistical randomness (uniform distribution) here. The distinction is that it might be possible to guess which trace id is generated in some cases (through things like timing attacks), but that's fine in this case.

Yes I did try crypto: 10x slower than the original 🫠

Well, apparently it's slower than your solution... this was unexpected 😄 https://jsben.ch/ZR73M

We were actually using crypto module in the past and removed it for performance reasons https://github.com/open-telemetry/opentelemetry-js/pull/1349/changes

I just added a benchmark test.

…owser-random-id

david-luna · 2025-12-18T08:39:45Z

packages/opentelemetry-sdk-trace-base/src/platform/browser/RandomIdGenerator.ts

+    buf[i] = (Math.random() * 256) >>> 0;
+  }
+  // Ensure non-zero
+  for (let i = 0; i < buf.length; i++) {


nit: as a fallback a '1' is written in the last position to ensure the buf does not contain all zeroes. Maybe accessing a random position of the array and setting it to '1' if not already set would allow is to skip a second iteration.

Maybe accessing a random position of the array and setting it to '1' if not already set would allow is to skip a second iteration.

I'm not a mathematician but I'd worry a bit that that biases the distribution away from values with 0s.

Me neither but what you said makes a lot of sense. The I guess a flag in the loop that fills the buf would be enough

Something like

let allZero = true; for (let i = 0; i < buf.length; i++) { buf[i] = (Math.random() * 256) >>> 0; allZero = allZero && buf[i] === 0; } if (allZero) { buf[buf.length - 1] = 1; }

@david-luna the allZero check in the loop is about 20% slower than the proposed function.

trentm

LGTM, with a Q about dropping the benchmark file.

trentm · 2025-12-18T20:15:37Z

packages/opentelemetry-sdk-trace-base/karma.bench.js

Do we keep this file? This is a benchmark running node using browser code. I think quoting some benchmark code (or perhaps a https://jsben.ch or similar link) showing browser numbers in the PR discussion would be good, but not sure of the value of having the benchmark file persisting in the repo.

Great point. Let me try to add a browser bench test.

@trentm I moved some files around so the platform chosen in package.json is used for common tests, including the new benchmark test for the randomIdGenerator. It expanded the scope quite a bit but now reports numbers for node and browser. I updated the description above with node, browser (main), and browser (this branch) results.

If this is too many changes for this PR I can back them out and make a new branch for the test folder.

@trentm I added browser benchmark tests that do not write to disk. I just saw @pichlermarc's note about being judicious with the amount of benchmarks published.

I defer to you whether or not they should be published (in a follow-up)

Great, thanks. Sounds good to me. My main concern had been about the (lack of) value of having those metrics included in the published benchmarks.

Marc had mentioned that he has a PR coming at some point to reduce the set of metrics that will be included in the published set (by just changing the benchmark to not output results to the benchmark.txt file that is slurped up by the benchmark.yml CI, I believe).

…owser-random-id

- move test/common/platform/RandomIdGenerator.test.ts to test/common/ - move test/common/Sampler.test.ts to test/common/sampler/ - add benchmarks for RandomIdGenerator, Span, and BatchSpanProcessor - benchmarks run on both Node and browser platforms - add karma.bench.js config for browser benchmarks - add AMD stub to karma.webpack.js for Benchmark.js compatibility

…owser-random-id

trentm

I think this LGTM. Impl is good. I didn't dig into the browser bench scripts.

trentm · 2026-01-21T19:08:07Z

packages/opentelemetry-sdk-trace-base/karma.bench.js

Great, thanks. Sounds good to me. My main concern had been about the (lack of) value of having those metrics included in the published benchmarks.

Marc had mentioned that he has a PR coming at some point to reduce the set of metrics that will be included in the published set (by just changing the benchmark to not output results to the benchmark.txt file that is slurped up by the benchmark.yml CI, I believe).

perf(sdk-trace-base): use Uint8Array for browser RandomIdGenerator

b3791ae

overbalance requested a review from a team as a code owner December 10, 2025 23:48

chore: add changelog entry for browser RandomIdGenerator perf

ee2d94d

BenoitZugmeyer reviewed Dec 11, 2025

View reviewed changes

overbalance requested a review from dyladan December 12, 2025 17:18

overbalance enabled auto-merge December 15, 2025 06:04

overbalance added 3 commits December 17, 2025 14:47

Merge remote-tracking branch 'upstream/main' into overbalance/perf-br…

d7fe798

…owser-random-id

test(sdk-trace-base): add benchmark for browser RandomIdGenerator

be94956

Merge remote-tracking branch 'upstream/main' into overbalance/perf-br…

462afae

…owser-random-id

overbalance added the browser Browser-specific additions or benefits label Dec 17, 2025

david-luna reviewed Dec 18, 2025

View reviewed changes

trentm approved these changes Dec 18, 2025

View reviewed changes

overbalance disabled auto-merge December 18, 2025 21:04

overbalance added 5 commits December 18, 2025 16:08

Merge branch 'main' into overbalance/perf-browser-random-id

a0b0a12

Merge remote-tracking branch 'upstream/main' into overbalance/perf-br…

b8a4740

…owser-random-id

test(sdk-trace-base): add browser benchmark for RandomIdGenerator

e04cec4

Update karma.webpack.js

16ca8c7

overbalance requested a review from trentm December 19, 2025 03:54

overbalance added 3 commits December 19, 2025 10:42

chore: dont include config in coverage

57e44ed

Merge branch 'main' into overbalance/perf-browser-random-id

359171b

Merge branch 'main' into overbalance/perf-browser-random-id

8f9588e

overbalance mentioned this pull request Jan 6, 2026

perf: optimize regex, remove uuid() embrace-io/embrace-web-sdk#1044

Merged

overbalance added 6 commits January 6, 2026 14:17

Merge remote-tracking branch 'upstream/main' into overbalance/perf-br…

b3ad1c7

…owser-random-id

Merge branch 'main' into overbalance/perf-browser-random-id

57076cf

Merge remote-tracking branch 'upstream/main' into overbalance/perf-br…

d7cadc9

…owser-random-id

Merge remote-tracking branch 'upstream/main' into overbalance/perf-br…

65d7875

…owser-random-id

Merge remote-tracking branch 'upstream/main' into overbalance/perf-br…

449ec34

…owser-random-id

Update CHANGELOG.md

1b2ff44

overbalance added 3 commits January 13, 2026 10:55

Merge remote-tracking branch 'upstream/main' into overbalance/perf-br…

06e9418

…owser-random-id

Merge branch 'main' into overbalance/perf-browser-random-id

c2d34a4

don't write browser benchmarks to log

6f0629b

overbalance force-pushed the overbalance/perf-browser-random-id branch 3 times, most recently from be43a55 to bb60e72 Compare January 20, 2026 21:25

refactor(sdk-trace-base): split benchmarks into node and browser

4c703e2

overbalance force-pushed the overbalance/perf-browser-random-id branch from bb60e72 to 4c703e2 Compare January 20, 2026 21:30

whitespace

46a2879

david-luna approved these changes Jan 21, 2026

View reviewed changes

overbalance added 4 commits January 21, 2026 12:46

Merge remote-tracking branch 'upstream/main' into overbalance/perf-br…

e0649b2

…owser-random-id

dont write browser benchmarks to disk

6cb1e9b

add global command

ebc8bbe

Update CHANGELOG.md

3499c8f

overbalance enabled auto-merge January 21, 2026 19:00

overbalance added this pull request to the merge queue Jan 21, 2026

Merged via the queue into open-telemetry:main with commit e5f2d42 Jan 21, 2026
27 checks passed

overbalance deleted the overbalance/perf-browser-random-id branch January 21, 2026 19:11

trentm approved these changes Jan 21, 2026

View reviewed changes

perf(sdk-trace-base): use Uint8Array for browser RandomIdGenerator #6209

perf(sdk-trace-base): use Uint8Array for browser RandomIdGenerator #6209

Uh oh!

Conversation

overbalance commented Dec 10, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Benchmark Results

Test plan

Benchmark Infrastructure

Separate Node and browser benchmarks

Output files

Test folder structure

Uh oh!

codecov bot commented Dec 10, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

overbalance Jan 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

trentm left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

overbalance Dec 19, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

overbalance Jan 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

trentm left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

overbalance commented Dec 10, 2025 •

edited

Loading

codecov bot commented Dec 10, 2025 •

edited

Loading

overbalance Jan 20, 2026 •

edited

Loading

overbalance Dec 19, 2025 •

edited

Loading

overbalance Jan 20, 2026 •

edited

Loading