perf: Make every gossip thread use its own randomness instance, reducing contention #3006
Conversation
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
ValarDragon
commented
May 5, 2024
Comment on lines
314
to
316
| // NOTE: This relies on the os's random number generator. | ||
| // For real security, we should salt that with some seed. | ||
| // See github.com/cometbft/cometbft/crypto for a more secure reader. |
There was a problem hiding this comment.
I disagree with this comment, but will leave to the general cleanup PR for this package to remove.
There was a problem hiding this comment.
This comment does not make sense given the preamble of this source file...
ValarDragon
commented
May 6, 2024
Comment on lines
+45
to
+46
| // G404: Use of weak random number generator (math/rand instead of crypto/rand) | ||
| //nolint:gosec |
There was a problem hiding this comment.
Note that the way we create the local Rand object also has this nolint, so its not newly introduced.
cason
approved these changes
May 7, 2024
There was a problem hiding this comment.
Using a synchronized random source here does not make sense.
The solution, however, is pretty complex, but it is what we can do right now. I would also check other internal uses of this internal random library to understand when we actually need a thread-safe version of it.
| @@ -87,6 +87,24 @@ func TestRngConcurrencySafety(_ *testing.T) { | |||
| wg.Wait() | |||
| } | |||
|
|
|||
| // Makes a new stdlib random instance 100 times concurrently. | |||
There was a problem hiding this comment.
Should we test that cRandBytes is thread safe instead?
Comment on lines
314
to
316
| // NOTE: This relies on the os's random number generator. | ||
| // For real security, we should salt that with some seed. | ||
| // See github.com/cometbft/cometbft/crypto for a more secure reader. |
There was a problem hiding this comment.
This comment does not make sense given the preamble of this source file...
There was a problem hiding this comment.
Somehow impressed that the bit array method is only used here. I wonder whether implementing part of this logic here, instead of there, could be a better solution.
At the end, we are counting the 1 bits, then getting a random index from zero to this number, then retrieving the i-th 1 bit. The random part we could implement outside the bit array library.
|
Every other usage from what I can tell can just be replaced with math.Rand, but wanted to do that in a separate PR. (Math.rand's global instance is thread safe, by internally doing CAS operations and a retry loop with no wait time) |
|
The problem of |
mergify bot
pushed a commit
that referenced
this pull request
May 9, 2024
…ing contention (#3006) Closes #3005 As noted in that issue, we currently are doing extra CPU overhead and mutex contention for getting a random number. This PR removes this overhead by making every performance sensitive thread have its own rand instance. In a subsequent PR, we can cleanup all the testing usages, and likely just entirely delete our internal randomness package. I didn't do that in this PR to keep it straightforward to verify. --- #### PR checklist - [x] Tests written/updated - [ ] Changelog entry added in `.changelog` (we use [unclog](https://github.com/informalsystems/unclog) to manage our changelog) - [ ] Updated relevant documentation (`docs/` or `spec/`) and code comments - [ ] Title follows the [Conventional Commits](https://www.conventionalcommits.org/en/v1.0.0/) spec (cherry picked from commit df28e73) # Conflicts: # internal/consensus/reactor.go
4 tasks
melekes
added a commit
that referenced
this pull request
May 9, 2024
…ing contention (backport #3006) (#3045) Closes #3005 As noted in that issue, we currently are doing extra CPU overhead and mutex contention for getting a random number. This PR removes this overhead by making every performance sensitive thread have its own rand instance. In a subsequent PR, we can cleanup all the testing usages, and likely just entirely delete our internal randomness package. I didn't do that in this PR to keep it straightforward to verify. --- #### PR checklist - [x] Tests written/updated - [ ] Changelog entry added in `.changelog` (we use [unclog](https://github.com/informalsystems/unclog) to manage our changelog) - [ ] Updated relevant documentation (`docs/` or `spec/`) and code comments - [ ] Title follows the [Conventional Commits](https://www.conventionalcommits.org/en/v1.0.0/) spec <hr>This is an automatic backport of pull request #3006 done by [Mergify](https://mergify.com). --------- Co-authored-by: Dev Ojha <ValarDragon@users.noreply.github.com> Co-authored-by: Anton Kaliaev <anton.kalyaev@gmail.com>
ValarDragon
added a commit
to osmosis-labs/cometbft
that referenced
this pull request
May 26, 2024
…ing contention (cometbft#3006) Closes cometbft#3005 As noted in that issue, we currently are doing extra CPU overhead and mutex contention for getting a random number. This PR removes this overhead by making every performance sensitive thread have its own rand instance. In a subsequent PR, we can cleanup all the testing usages, and likely just entirely delete our internal randomness package. I didn't do that in this PR to keep it straightforward to verify. --- - [x] Tests written/updated - [ ] Changelog entry added in `.changelog` (we use [unclog](https://github.com/informalsystems/unclog) to manage our changelog) - [ ] Updated relevant documentation (`docs/` or `spec/`) and code comments - [ ] Title follows the [Conventional Commits](https://www.conventionalcommits.org/en/v1.0.0/) spec
Merged
7 tasks
mergify bot
pushed a commit
to osmosis-labs/cometbft
that referenced
this pull request
May 28, 2024
…ing contention (cometbft#3006) Closes cometbft#3005 As noted in that issue, we currently are doing extra CPU overhead and mutex contention for getting a random number. This PR removes this overhead by making every performance sensitive thread have its own rand instance. In a subsequent PR, we can cleanup all the testing usages, and likely just entirely delete our internal randomness package. I didn't do that in this PR to keep it straightforward to verify. --- - [x] Tests written/updated - [ ] Changelog entry added in `.changelog` (we use [unclog](https://github.com/informalsystems/unclog) to manage our changelog) - [ ] Updated relevant documentation (`docs/` or `spec/`) and code comments - [ ] Title follows the [Conventional Commits](https://www.conventionalcommits.org/en/v1.0.0/) spec (cherry picked from commit f55b9f4)
7 tasks
PaddyMc
added a commit
to osmosis-labs/cometbft
that referenced
this pull request
May 28, 2024
…… (backport #77) (#82) * perf: Make every gossip thread use its own randomness instance, reducing contention (cometbft#3006) Closes cometbft#3005 As noted in that issue, we currently are doing extra CPU overhead and mutex contention for getting a random number. This PR removes this overhead by making every performance sensitive thread have its own rand instance. In a subsequent PR, we can cleanup all the testing usages, and likely just entirely delete our internal randomness package. I didn't do that in this PR to keep it straightforward to verify. --- - [x] Tests written/updated - [ ] Changelog entry added in `.changelog` (we use [unclog](https://github.com/informalsystems/unclog) to manage our changelog) - [ ] Updated relevant documentation (`docs/` or `spec/`) and code comments - [ ] Title follows the [Conventional Commits](https://www.conventionalcommits.org/en/v1.0.0/) spec (cherry picked from commit f55b9f4) * Add Changelgo (cherry picked from commit ce04f04) # Conflicts: # CHANGELOG.md * Fix changelog further (cherry picked from commit bd34ce6) # Conflicts: # CHANGELOG.md --------- Co-authored-by: Dev Ojha <ValarDragon@users.noreply.github.com> Co-authored-by: Dev Ojha <dojha@berkeley.edu> Co-authored-by: PaddyMc <paddymchale@hotmail.com>
PaddyMc
added a commit
to osmosis-labs/cometbft
that referenced
this pull request
Aug 19, 2024
…… (backport #77) (#82) * perf: Make every gossip thread use its own randomness instance, reducing contention (cometbft#3006) Closes cometbft#3005 As noted in that issue, we currently are doing extra CPU overhead and mutex contention for getting a random number. This PR removes this overhead by making every performance sensitive thread have its own rand instance. In a subsequent PR, we can cleanup all the testing usages, and likely just entirely delete our internal randomness package. I didn't do that in this PR to keep it straightforward to verify. --- - [x] Tests written/updated - [ ] Changelog entry added in `.changelog` (we use [unclog](https://github.com/informalsystems/unclog) to manage our changelog) - [ ] Updated relevant documentation (`docs/` or `spec/`) and code comments - [ ] Title follows the [Conventional Commits](https://www.conventionalcommits.org/en/v1.0.0/) spec (cherry picked from commit f55b9f4) * Add Changelgo (cherry picked from commit ce04f04) * Fix changelog further (cherry picked from commit bd34ce6) --------- Co-authored-by: Dev Ojha <ValarDragon@users.noreply.github.com> Co-authored-by: Dev Ojha <dojha@berkeley.edu> Co-authored-by: PaddyMc <paddymchale@hotmail.com>
Merged
7 tasks
ValarDragon
added a commit
to osmosis-labs/cometbft
that referenced
this pull request
Aug 19, 2024
…… (backport #77) (#82) (#132) * perf: Make every gossip thread use its own randomness instance, reducing contention (cometbft#3006) Closes cometbft#3005 As noted in that issue, we currently are doing extra CPU overhead and mutex contention for getting a random number. This PR removes this overhead by making every performance sensitive thread have its own rand instance. In a subsequent PR, we can cleanup all the testing usages, and likely just entirely delete our internal randomness package. I didn't do that in this PR to keep it straightforward to verify. --- - [x] Tests written/updated - [ ] Changelog entry added in `.changelog` (we use [unclog](https://github.com/informalsystems/unclog) to manage our changelog) - [ ] Updated relevant documentation (`docs/` or `spec/`) and code comments - [ ] Title follows the [Conventional Commits](https://www.conventionalcommits.org/en/v1.0.0/) spec (cherry picked from commit f55b9f4) * Add Changelgo (cherry picked from commit ce04f04) * Fix changelog further (cherry picked from commit bd34ce6) --------- Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> Co-authored-by: Dev Ojha <ValarDragon@users.noreply.github.com> Co-authored-by: Dev Ojha <dojha@berkeley.edu>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Closes #3005
As noted in that issue, we currently are doing extra CPU overhead and mutex contention for getting a random number. This PR removes this overhead by making every performance sensitive thread have its own rand instance.
In a subsequent PR, we can cleanup all the testing usages, and likely just entirely delete our internal randomness package. I didn't do that in this PR to keep it straightforward to verify.
PR checklist
.changelog(we use unclog to manage our changelog)docs/orspec/) and code comments