PolarDB:PolarSearch 1.0 performance test

term: Exact term match

An exact term match is performed on logs-* with a fixed hit count of 10,000 entries, returning the default 10 entries. This test evaluates the search efficiency of the PolarSearch 1.0 inverted index, the overhead of cross-shard fan-out and merge operations, and the processing capability of the network coordination layer.

asc_sort_timestamp: Sorting by doc values

A match_all query is executed on all documents, and the top 10 results are returned after being sorted in ascending order by timestamp. The @timestamp field is of the date type and is sorted by using doc values columnar storage. This operation evaluates the read efficiency of doc values, top-N heap sorting, and the merging capability of the coordinator node for PolarSearch 1.0.

range: Time range scan

Perform a dynamic time window range query on the @timestamp field. Because 90% of queries in the log product include a time filter, this task evaluates the efficiency of date range scanning, complementing term queries to provide a comparison between point queries and range queries.

hourly_agg: Time-series aggregation

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

match-all: Baseline scheduling

A match_all query is executed on the entire dataset of approximately 165 million records, returning only the default 10 records without sorting or aggregation. This process tests the efficiency of the basic path for engine request parsing, shard scheduling, and result serialization.

range: Numeric range scan

Performs a fixed numeric range filter on the total_amount field (fare amount, scaled_float type) to query for trips with a fare between 5 and 15 dollars, and returns 10 results by default. This operation tests the range scanning efficiency of a numeric field.

autohisto_agg: Adaptive time aggregation

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

term: Exact term match

phrase: Full-text phrase match

This test evaluates phrase match performance in full-text search. A phrase match requires verifying term position information in the inverted index and is therefore more computationally complex than a standard term query.

country_agg_uncached: Terms aggregation with nesting

This test performs a terms aggregation (bucketing by country code) with a nested sum aggregation across all documents. The aggregation cache is disabled to reflect the true computational overhead. This test evaluates the engine's CPU utilization efficiency under heavy aggregation loads, making it representative of compute-intensive aggregation scenarios.

decay_geo_gauss_function_score: Geospatial distance decay scoring

A scoring query with a geo-distance decay function (Gaussian decay) is run on the geo_point field to evaluate the query efficiency of the geospatial index and the score calculation capability.

painless_static: Painless script scoring

This test uses a Painless script to compute a custom score for each document. The script is pre-compiled and executed statically (static binding). This test evaluates the computational throughput of the script engine, including JIT compilation. It is representative of highly customized script scoring scenarios.

desc_sort_population: Numeric sorting by doc values

Sort all documents by the population field (integer type) in descending order and return the top 10 results. The sorting is performed by using doc values columnar storage. This operation tests the efficiency of reading numeric fields from columnar storage and the performance of top-N sorting.