feat(perf): Add comprehensive benchmarking framework #3029
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
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: XuanYang-cn The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
…nsive benchmarking This commit includes comprehensive benchmark suite to identify remaining bottlenecks **New Benchmarks** Created comprehensive benchmark suite under `tests/benchmark/`: - **Access patterns**: 23 tests measuring real-world usage patterns - First-access (UI display), iterate-all (export), random-access (pagination) - **Search benchmarks**: Various vector types, dimensions, output fields - **Query benchmarks**: Scalars, JSON, all output types - **Hybrid search**: Multiple requests, varying top-k **Profiling Infrastructure** - Mock framework for client-only testing (no server required) - Integrated `pytest-memray` for memory profiling - Added helper scripts: - `profile_cpu.sh`: CPU profiling with py-spy - `profile_memory.sh`: Memory profiling with pytest-memray **Profiling Tools** - `pytest-benchmark`: Timing measurements - `py-spy`: CPU profiling and flamegraphs - `memray`: Memory allocation tracking **Key Discoveries** 1. **Lazy loading inefficiency** (CRITICAL) - Accessing first result materializes ALL results (+77% overhead) - Example: `result[0][0]` loads all 10,000 results - Impact: 423ms → 749ms for 10K results 2. **Vector materialization dominates** (HIGH PRIORITY) - 76% of memory usage (326 MiB of 431 MiB for 65K results) - 8x slower than scalars (337ms vs 42ms for 10K results) - Scales linearly with dimensions (128d: 8 MiB, 1536d: 68 MiB) 3. **Struct fields are slow** (MEDIUM PRIORITY) - 10x slower than scalars (435ms vs 42ms for 10K results) - Column-to-row conversion overhead - Linear O(n) scaling with high constant factor 4. **Scalars are efficient** (NO OPTIMIZATION NEEDED) - 64.6 MiB for 65K rows × 4 fields - ~1 KB per entity (acceptable dict overhead) Signed-off-by: yangxuan <[email protected]>
XuanYang-cn
force-pushed
the
bench-bench
branch
from
1a5ab37 to
6961059
Compare
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## master #3029 +/- ##
=========================================
Coverage ? 46.41%
=========================================
Files ? 65
Lines ? 13614
Branches ? 0
=========================================
Hits ? 6319
Misses ? 7295
Partials ? 0 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
XuanYang-cn
changed the title
XuanYang-cn
changed the title
XuanYang-cn
added
the
PR | need to cherry-pick to 2.x
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.
This commit includes comprehensive benchmark suite to identify remaining bottlenecks
New Benchmarks
Created comprehensive benchmark suite under
tests/benchmark/:- Access patterns: 23 tests measuring real-world usage patterns
- First-access (UI display), iterate-all (export), random-access (pagination)
- Search benchmarks: Various vector types, dimensions, output fields
- Query benchmarks: Scalars, JSON, all output types
- Hybrid search: Multiple requests, varying top-k
Profiling Infrastructure
- Mock framework for client-only testing (no server required)
- Integrated
pytest-memrayfor memory profiling- Added helper scripts:
-
profile_cpu.sh: CPU profiling with py-spy-
profile_memory.sh: Memory profiling with pytest-memrayProfiling Tools
-
pytest-benchmark: Timing measurements-
py-spy: CPU profiling and flamegraphs-
memray: Memory allocation trackingKey Discoveries
Signed-off-by: yangxuan [email protected]