Add Lucene specific file extensions to core HybridFS

[martin-gaievski]

Signed-off-by: Martin Gaievski [email protected]

Description

We're including file extensions for vector value files from Lucene 9.4 to the list of extensions that Core OpenSearch will use with HybridFS store type and MMap file I/O. This increases performance for both data ingestion and queries for p99. Setting is abstracted at engine level with specific implementation for Lucene. Setting is set at cluster defaults level, index specific overrides will take priority over it.

Tested locally on 1M dataset

Before change (baseline):
data ingestion

  "results": {
    "test_took": 2706990.31884,
    "create_index_took_total": 2702.2581160000004,
    "ingest_took_total": 513438.04862599995,
    "refresh_index_store_kb_total": 2810943.876953125,
    "refresh_index_took_total": 641.9670379999616,
    "force_merge_took_total": 2190208.04506
  }

query

  "results": {
    "test_took": 347624.7275845,
    "clear_cache_took_total": 22.227584499987294,
    "query_took_total": 347602.5,
    "query_took_p50": 32.5,
    "query_took_p90": 45.5,
    "query_took_p99": 83.0,
    "query_memory_kb_total": 0.0,
    "query_recall@K_total": 0.991061,
    "query_recall@1_total": 1.0
  }

With the change
data ingestion

  "results": {
    "test_took": 748055.190353,
    "create_index_took_total": 2722.189422,
    "ingest_took_total": 535081.281891,
    "refresh_index_store_kb_total": 2814413.49609375,
    "refresh_index_took_total": 1110.5574829999796,
    "force_merge_took_total": 209141.16155700004
  }

query

  "results": {
    "test_took": 111353.221852,
    "clear_cache_took_total": 110.22185199999768,
    "query_took_total": 111243.0,
    "query_took_p50": 9.5,
    "query_took_p90": 16.0,
    "query_took_p99": 28.5,
    "query_memory_kb_total": 0.0,
    "query_recall@K_total": 0.990751,
    "query_recall@1_total": 1.0
  }

Issues Resolved

#637

Check List

• New functionality includes testing.
  • All tests pass
• New functionality has javadoc added
• Commits are signed as per the DCO using --signoff

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.

[navneet1v]

Don't we want to load the files early using something like this: https://github.com/opensearch-project/OpenSearch/blob/2.5/server/src/main/java/org/opensearch/index/IndexModule.java#L147

[navneet1v]

Overall code looks good to me. Please reply comment of Jack on the adding another file type.

[codecov-commenter]

Codecov Report

Merging #721 (1c81f29) into main (5f4bef9) will decrease coverage by 0.21%.
The diff coverage is 100.00%.

@@             Coverage Diff              @@
##               main     #721      +/-   ##
============================================
- Coverage     84.64%   84.43%   -0.22%     
  Complexity     1072     1072              
============================================
  Files           151      152       +1     
  Lines          4345     4356      +11     
  Branches        389      389              
============================================
  Hits           3678     3678              
- Misses          489      498       +9     
- Partials        178      180       +2     

Impacted Files	Coverage Δ
.../java/org/opensearch/knn/index/util/KNNEngine.java	100.00% <100.00%> (ø)
...java/org/opensearch/knn/index/util/KNNLibrary.java	100.00% <100.00%> (ø)
...ain/java/org/opensearch/knn/index/util/Lucene.java	93.75% <100.00%> (+0.41%)	⬆️
...main/java/org/opensearch/knn/plugin/KNNPlugin.java	100.00% <100.00%> (ø)
...va/org/opensearch/knn/index/KNNCircuitBreaker.java	60.00% <0.00%> (-20.00%)	⬇️
...ain/java/org/opensearch/knn/index/KNNSettings.java	80.88% <0.00%> (-2.21%)	⬇️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.