[Indexing] Optimize Bloom Filter implementation to reduce locking #17167

mgodwan · 2025-01-28T20:27:28Z

Is your feature request related to a problem? Please describe

Improve implementation of bloom filter. This may involve the need to create some new metadata while initializing the FieldsProducer

Describe the solution you'd like

Initialize different bloom filter instances for different queries rather than using a single one which has synchronized block today on the long array trying to read the byte values

OpenSearch/server/src/main/java/org/opensearch/index/codec/fuzzy/IndexInputImmutableLongArray.java

Line 41 in 32a88eb

public synchronized long get(long index) {

Related component

Indexing:Performance

Describe alternatives you've considered

No response

Additional context

No response

mgodwan added enhancement Enhancement or improvement to existing feature or request untriaged labels Jan 28, 2025

mgodwan self-assigned this Jan 28, 2025

github-actions bot added the Indexing:Performance label Jan 28, 2025

mgodwan removed the untriaged label Jan 30, 2025

mgodwan linked a pull request Jan 30, 2025 that will close this issue

[Optimization] Move bloom filter init logic outside of the FuzzyFilteredFieldsProducer constructor #17200

Draft

3 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[Indexing] Optimize Bloom Filter implementation to reduce locking #17167

[Indexing] Optimize Bloom Filter implementation to reduce locking #17167

mgodwan commented Jan 28, 2025 •

edited

Loading

[Indexing] Optimize Bloom Filter implementation to reduce locking #17167

[Indexing] Optimize Bloom Filter implementation to reduce locking #17167

Comments

mgodwan commented Jan 28, 2025 • edited Loading

Is your feature request related to a problem? Please describe

Describe the solution you'd like

Related component

Describe alternatives you've considered

Additional context

mgodwan commented Jan 28, 2025 •

edited

Loading