PolarDB:PolarSearch 3.0 performance

term - Term match

This test performs a term match on logs-* indices. Each query hits a fixed 10,000 documents and returns the default 10 documents. The test measures the efficiency of inverted index lookups, the overhead of cross-shard fan-out and merge operations, and the processing capacity of the coordinating node in PolarSearch 3.0.

asc_sort_timestamp - Doc value sorting

This test runs a match_all query on all documents, sorts them by timestamp in ascending order, and returns the top 10. The @timestamp field, a date type, is sorted using doc values, which rely on columnar storage. This test evaluates the efficiency of reading doc values, performing top-N heap sorts, and merging results on the coordinating node for PolarSearch 3.0.

range - Time range query

This test performs a dynamic time-window range query on the @timestamp field. Because 90% of queries in log analytics products include a time filter, this task measures the efficiency of date range queries. It serves as a counterpart to the term test, comparing range query performance against point query performance.

hourly_agg - Time-series aggregation

This test performs a date_histogram aggregation, bucketing by the hour across all documents. It measures the aggregation throughput for date types, a core operation for log monitoring and statistical analysis scenarios.

match-all - Baseline dispatching

This test runs a match_all query on approximately 165 million records, returning only the default 10 results without any sorting or aggregation. It measures the baseline efficiency of the engine's request parsing, shard dispatching, and result serialization.

range - Numeric range query

This test applies a fixed numeric range filter to the total_amount field (a scaled_float type representing the fare amount). It finds trips with fares between 5 and 15 dollars, returning the default 10 results. The test evaluates the efficiency of numeric field range queries.

autohisto_agg - Adaptive time aggregation

Perform auto_date_histogram (automatic bucketing, buckets=20) on data within a 20-day time window to test the engine's adaptive aggregation capability. This operation complements range (range is for numeric range queries, while auto_date_histogram is for time range aggregations).

term - Term match

phrase - Phrase query

This test evaluates phrase query capabilities within full-text search. It requires verifying term position information in the inverted index, which makes it more computationally complex than a term query.

country_agg_uncached - Terms and nested aggregation

This test performs a terms aggregation (bucketing by country code) with a nested sum aggregation on all documents. The aggregation cache is disabled to reflect the true computational overhead. This task represents a compute-intensive aggregation scenario and measures the engine's CPU efficiency under heavy aggregation loads.

decay_geo_gauss_function_score - Geospatial distance decay scoring

This test runs a function score query using Gaussian distance decay on the geo_point field. It evaluates both geospatial index query efficiency and scoring performance.

painless_static - Painless custom scoring

This test uses a Painless script to perform custom score calculations on each document. The script is pre-compiled and executed using static binding. This task measures the computational throughput of the script engine (JIT compilation) and represents a typical scenario for highly customized script scoring.

desc_sort_population - Numeric doc value sorting

This test sorts all documents in descending order by the population field (an integer type) and returns the top 10 results. Sorting is performed using doc values, which rely on columnar storage. The test evaluates the efficiency of reading columnar numeric fields and the performance of top-N sorting.