ChrisHegarty commented on PR #14980:
URL: https://github.com/apache/lucene/pull/14980#issuecomment-3141192639

   Some luceneutil benchmark results, showing a nice improvement in search 
latency and force merge times.
   
   Linux Skylake
   ```
   baseline 
   recall  latency(ms)  netCPU  avgCpuCount     nDoc  topK  fanout  maxConn  
beamWidth  quantized  index(s)  index_docs/s  force_merge(s)  num_segments  
index_size(MB)  vec_disk(MB)  vec_RAM(MB)  indexType
   0.942        2.502   2.497        0.998  1000000   100      50       64      
  250         no   2021.03        494.80          959.16             1         
3022.71      2929.688     2929.688       HNSW
   0.942        2.528   2.521        0.997  1000000   100      50       64      
  250         no      0.00      Infinity            0.09             1         
3022.71      2929.688     2929.688       HNSW
   0.942        2.525   2.518        0.997  1000000   100      50       64      
  250         no      0.00      Infinity            0.09             1         
3022.71      2929.688     2929.688       HNSW
   
   candidate
   0.941        2.029   2.016        0.994  1000000   100      50       64      
  250         no   1969.44        507.76          572.05             1         
3020.53      2929.688     2929.688       HNSW
   0.941        2.019   2.011        0.996  1000000   100      50       64      
  250         no      0.00      Infinity            0.09             1         
3020.53      2929.688     2929.688       HNSW
   0.941        2.019   2.011        0.996  1000000   100      50       64      
  250         no      0.00      Infinity            0.09             1         
3020.53      2929.688     2929.688       HNSW
   ```
   
   linux-arm (c6g.8xlarge)
   ```
   baseline (forcemerge)
   recall  latency(ms)  netCPU  avgCpuCount    nDoc  topK  fanout  maxConn  
beamWidth  quantized  index(s)  index_docs/s  force_merge(s)  num_segments  
index_size(MB)  vec_disk(MB)  vec_RAM(MB)  indexType
    0.939        4.012   4.003        0.998  500000   100      50       32      
  250         no    300.48       1664.02          511.45             1         
1504.30      1464.844     1464.844       HNSW
    0.939        4.139   4.130        0.998  500000   100      50       32      
  250         no      0.00      Infinity            0.16             1         
1504.30      1464.844     1464.844       HNSW
    0.939        4.207   4.198        0.998  500000   100      50       32      
  250         no      0.00      Infinity            0.17             1         
1504.30      1464.844     1464.844       HNSW
   candidate
    0.936        2.895   2.881        0.995  500000   100      50       32      
  250         no    221.61       2256.25          325.71             1         
1503.43      1464.844     1464.844       HNSW
    0.936        2.884   2.875        0.997  500000   100      50       32      
  250         no      0.00      Infinity            0.18             1         
1503.43      1464.844     1464.844       HNSW
    0.936        2.845   2.836        0.997  500000   100      50       32      
  250         no      0.00      Infinity            0.17             1         
1503.43      1464.844     1464.844       HNSW
   ```


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org

For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org

Reply via email to