As part of performance optimization for Hybrid query we need to find a way to minimize time taken by parallel score collecting processes that is running in core with TopDocsCollector. As per following information collected during profiling these calls take 40-80% of CPU time.
As a baseline we're taking results from previous PR related to hybrid query optimization, those are based on 2.13 version and noaa OSB workload, all times are in ms:
One sub-query that selects 11M documents
Bool: p50 97.9306 | p90 116.299
Hybrid: p50 228.696 | p90 249.665
One sub-query that selects 1.6K documents
Bool: p50 87.3152 | p90 89.3061
Hybrid: p50 89.9654 | p90 92.349
Three sub-query that select 15M documents
Bool: p50 97.9891 | p90 114.396
Hybrid: p50 353.631 | p90 377.527
Most likely that will be a compound change in both core OpenSearch and neural-search plugin. Preferred suggestion is - provide capability to skip or ignore TopDocsCollector in core QueryPhaseSearcher (Core side) and by using that new option call only HybridQueryDocsCollector (plugin side).
As part of performance optimization for Hybrid query we need to find a way to minimize time taken by parallel score collecting processes that is running in core with TopDocsCollector. As per following information collected during profiling these calls take 40-80% of CPU time.
As a baseline we're taking results from previous PR related to hybrid query optimization, those are based on 2.13 version and noaa OSB workload, all times are in ms:
Most likely that will be a compound change in both core OpenSearch and neural-search plugin. Preferred suggestion is - provide capability to skip or ignore TopDocsCollector in core QueryPhaseSearcher (Core side) and by using that new option call only HybridQueryDocsCollector (plugin side).