"The Numbers Query Latency (P50 / P95 / P99) Query Type PostgreSQL Elasticsearch Exact name match 180ms / 420ms / 890ms 8ms /"
A national platform serving 1M+ users needed search across citizen profiles, training courses, and job listings. Peak ingestion included millions of profiles, hundreds of thousands of courses, and job listings, with 50,000–80,000 queries per day and spikes to 8,000 queries per hour. Searches required partial matches, typo tolerance, relevance ranking, and faceted filtering across names, skills, locations, courses, and job categories. Both PostgreSQL 14.x with pg_trgm and tsvector and Elasticsearch 8.x were benchmarked on identical AWS t3.xlarge instances using representative queries and Apache JMeter load tests. Results showed PostgreSQL query latency far higher than Elasticsearch, especially at higher percentiles.
#search-infrastructure #postgresql-full-text-search #elasticsearch #performance-benchmarking #typo-tolerant-relevance
Read at Medium
Unable to calculate read time
Collection
[
|
...
]