Performance Optimization - Maximum Search Speed for 55M Vector Collection

[dmitriivahrushev]

Hello everyone!!!

Could you please advise how to achieve maximum search speed for a 55 million vector collection while maintaining 99% precision? We have a 10-node cluster with 265GiB RAM per node.

Current performance: approximately 30 RPS with 16 concurrent threads.

Collection Configuration:

{
  "status": "green",
  "optimizer_status": "ok",
  "indexed_vectors_count": 55046191,
  "points_count": 55085603,
  "segments_count": 75,
  "config": {
    "params": {
      "vectors": {
        "size": 512,
        "distance": "Cosine",
        "hnsw_config": {
          "m": 16,
          "ef_construct": 1024
        },
        "datatype": "float16"
      },
      "shard_number": 10,
      "replication_factor": 1,
      "write_consistency_factor": 1,
      "on_disk_payload": true
    },
    "hnsw_config": {
      "m": 16,
      "ef_construct": 100,
      "full_scan_threshold": 10000,
      "max_indexing_threads": 0,
      "on_disk": false
    },
    "optimizer_config": {
      "deleted_threshold": 0.2,
      "vacuum_min_vector_number": 1000,
      "default_segment_number": 0,
      "max_segment_size": null,
      "memmap_threshold": null,
      "indexing_threshold": 10000,
      "flush_interval_sec": 5,
      "max_optimization_threads": null
    },
    "wal_config": {
      "wal_capacity_mb": 32,
      "wal_segments_ahead": 0,
      "wal_retain_closed": 1
    },
    "quantization_config": null
  }
}

Search Implementation:
Vector search is implemented via batch search through API with batches sent across 16 separate threads:
endpoint = f"collections/{collection_name}/points/query/batch"

What optimizations would you recommend to improve search performance from the current 30 RPS?

Thank you!