[PERF]: Improved server-side serialization with orjson + async #1938
Conversation
This file contains 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
Reviewer ChecklistPlease leverage this checklist to ensure your code review is thorough before approving Testing, Bugs, Errors, Logs, Documentation
System Compatibility
Quality
|
|
Picking up from #1695 (comment):
|
2 tasks
- Added orjson serialization improving serialization performance especially on large batches 100+ docs 2x faster (tested with locust) - Added async body serialization further improving performance (tested with locust) - Added async handling with AnyIO of the more impactful server queries further reducing concurrent request response times Future work: Add orjson serialization at client-side
…PI Server thread pool connections
- Removed commented out code - Added clarification for to_thread usage
- Configurable via `chroma_server_thread_pool_size` setting
…f the main loop TODO: Telemetry/Authz has the potentially to negatively impact things - needs async refactor
tazarov
force-pushed
the
feature/orjson-parser-async-endpoints
branch
from
76492a2 to
ed53f6a
Compare
- Body serialization is off the main thread
- Get collection did not have enough args - Body was not awaited in modify collection - Minor cleanups to make code more readable
beggers
reviewed
Apr 2, 2024
| @@ -789,7 +791,7 @@ def reset(self) -> bool: | |||
| def get_settings(self) -> Settings: | |||
| return self._settings | |||
|
|
|||
| @property | |||
| @cached_property | |||
There was a problem hiding this comment.
is there a reason for this? it adds a small amount of cognitive overhead ("what's a cached property?") for no benefit that i can see?
| @@ -34,6 +34,7 @@ def __init__(self, conn_pool: Pool, stack: local): | |||
| @override | |||
| def __enter__(self) -> base.Cursor: | |||
| if len(self._tx_stack.stack) == 0: | |||
| self._conn.execute("PRAGMA case_sensitive_like = ON") | |||
There was a problem hiding this comment.
Was this meant to be included? It seems unrelated to the rest of the changes.
There was a problem hiding this comment.
I found this bug while testing the PR, so I added it here. I can split it up in another PR, as it is indeed unrelated.
There was a problem hiding this comment.
Move to PR or leave it here?
|
|
||
| return wrapper | ||
| if asyncio.iscoroutinefunction(f): | ||
|
|
There was a problem hiding this comment.
Can we fix the spacing here? It's funky to have an empty line after an if-statement
There was a problem hiding this comment.
Interesting, I am sure I have fixed this and ran pre-commit run which must have resulted in this formatting again.
beggers
reviewed
Apr 2, 2024
beggers
approved these changes
Apr 3, 2024
Co-authored-by: Ben Eggers <[email protected]>
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.
This is PR #1695 migrated to Chroma repo for stacking the outstanding changes.
Description of changes
Summarize the changes made by this PR.
Test plan
How are these changes tested?
pytestfor python,yarn testfor jsFor performance tests, 5 concurrent clients were querying, and 1 was adding pre-computed OpenAI (ada-02) embeddings (runtime ~1m)
Batch size: 100 (1000 also attached in a zip)
Tests with existing
maincodebase:Tests with orjson + async:
1m-100batch-5concur.zip
1m-1000batch-5concur.zip
Documentation Changes
Are all docstrings for user-facing APIs updated if required? Do we need to make documentation changes in the docs repository?
Refs