Elasticsearch vs PostgreSQL

This is not an either/or choice for most systems: Elasticsearch excels at search while PostgreSQL excels as a primary data store. Understanding when you need both versus when PostgreSQL alone suffices is key.

Architecture Differences

Elasticsearch stores data in inverted indexes optimized for full-text search. Every field is indexed by default, and documents are distributed across shards for parallel query execution. The Lucene-based engine provides BM25 relevance scoring, fuzzy matching, and nested aggregations.

PostgreSQL's full-text search uses tsvector (tokenized document) and tsquery (search query) types with GIN indexes. It is built into the database, requires no additional infrastructure, and provides respectable search capabilities for moderate workloads.

Performance Characteristics

For complex search queries with relevance ranking across millions of documents, Elasticsearch is significantly faster and more feature-rich. It handles fuzzy matching, synonym expansion, phrase proximity, and multi-field boosting natively.

PostgreSQL's full-text search is sufficient for basic keyword matching, phrase search, and simple ranking. For applications with moderate search needs (under a few million documents), PostgreSQL search eliminates the need for a separate search system.

Trade-offs

Adding Elasticsearch means maintaining a separate cluster, syncing data from your primary database, and handling eventual consistency between the source of truth and the search index. This adds operational complexity and introduces a sync lag.

For system design interviews, the common pattern is: PostgreSQL as the primary data store with Elasticsearch as a search index, synchronized via change data capture (CDC) or application-level dual writes.

When PostgreSQL Search Is Enough

If your application needs basic search over a moderate dataset with SQL-level consistency, PostgreSQL's built-in search is often sufficient. Many applications add Elasticsearch prematurely. Start with PostgreSQL search and migrate to Elasticsearch when you hit its limits.

Real-World Architecture

GitHub uses Elasticsearch for code search alongside PostgreSQL for data storage. Stack Overflow uses Elasticsearch for search but SQL Server for data. The dual-database pattern is standard for search-heavy applications.

Learn about search architecture patterns and data synchronization. See our system design examples and pricing.