Files
seaweedfs/weed
chrislu e76f632907 Phase 1: Add smart prefetching foundation for ML workloads
- Implement PrefetchManager with configurable worker pool and deduplication
- Add AccessPatternDetector for sequential, strided, and ML-specific patterns
- Create MLReaderCache with ML-aware prefetching capabilities
- Add comprehensive unit tests for prefetch manager
- Include foundation for detecting training datasets, model loading, and epoch patterns
- Support configurable prefetch parameters optimized for ML workloads

Features:
- Concurrent prefetch workers (8 by default)
- Pattern detection for sequential, model, epoch, and strided access
- ML-specific heuristics for large file and dataset access
- Comprehensive metrics and monitoring
- Graceful shutdown and cleanup

Tests:
- PrefetchManager: All tests passing (9/9)
- AccessPatternDetector: Core functionality implemented
- MLReaderCache: Basic functionality and integration tests
2025-08-30 15:04:36 -07:00
..
2025-08-09 21:47:29 -07:00
2025-07-19 23:56:17 -07:00
2025-08-30 11:15:48 -07:00
2025-08-30 11:16:26 -07:00
2025-08-30 11:15:48 -07:00
2025-08-22 01:15:42 -07:00
2025-08-22 01:15:42 -07:00
2025-08-30 11:15:48 -07:00
2024-06-25 09:18:11 -07:00
2025-08-30 11:15:48 -07:00
2025-08-30 11:15:48 -07:00
2024-02-14 08:26:38 -08:00
2025-08-30 11:15:48 -07:00
2025-07-19 21:43:34 -07:00
2025-03-17 23:13:27 -07:00