via “segment-based storage with automatic compaction and optimization”
Qdrant - High-performance, massive-scale Vector Database and Vector Search Engine for the next generation of AI. Also available in the cloud https://cloud.qdrant.io/
Unique: Implements segment-based storage with automatic compaction triggered by heuristics (segment size, fragmentation ratio) rather than manual thresholds, and integrates compaction into the segment lifecycle so HNSW indices are rebuilt during compaction rather than requiring separate index maintenance
vs others: More efficient than LSM-tree approaches because segments are optimized for vector search (columnar layout, HNSW indices) rather than generic key-value storage, and compaction is integrated with index building rather than separate