SQL Optimization Patterns

Specialist in diagnosing and fixing SQL performance bottlenecks through query plan analysis and indexing strategies.

When to Use This Skill

Use when:

• SQL queries exceed SLA or performance requirements
• Need to diagnose slow queries using EXPLAIN analysis
• Designing database schemas and indexing strategies for analytics workloads
• Reducing database costs through query efficiency
• Optimizing joins, aggregations, or complex analytical queries
• Investigating full table scans and expensive operations
• Implementing index strategies for high-performance reads
• Performance debugging when queries are slow in production

Core Capabilities

1. Query Plan Analysis - Read EXPLAIN output to identify bottlenecks
2. Indexing Strategies - Design B-tree, hash, GIN, BRIN indexes for query patterns
3. Join Optimization - Choose strategies (nested loop, hash, merge) based on data size
4. Schema Design - Denormalization, partitioning, materialized views for analytics
5. Query Refactoring - Rewrite slow queries using window functions, CTEs, batch operations
6. Cost Reduction - Improve query efficiency to reduce compute costs and resource usage

Reference Guides

For detailed implementation guidance, see:

EXPLAIN & Query Plans

Use when: Analyzing slow query performance

Covers:

• Understanding EXPLAIN output across PostgreSQL, MySQL, Snowflake, BigQuery
• Key metrics (Seq Scan, Index Scan, cost, rows, execution time)
• Reading execution plans from bottom-up
• Identifying bottlenecks and cardinality issues

Indexing Strategies

Use when: Designing indexes for analytics workloads

Covers:

• Index types (B-tree, hash, GIN, BRIN, covering, partial)
• Selectivity and composite index design
• Partial and expression indexes
• Avoiding over-indexing and unused indexes
• Index maintenance (ANALYZE, VACUUM, reindex)

Query Optimization Patterns

Use when: Rewriting queries for better performance

Covers:

• Eliminating N+1 query problems
• Cursor-based pagination (vs OFFSET)
• Optimizing COUNT and GROUP BY
• Transforming correlated subqueries
• Using CTEs and window functions
• Batch processing patterns

Advanced Techniques

Use when: Using advanced database features

Covers:

• Materialized views for pre-computation
• Table partitioning for large tables
• Query hints and optimization control
• Performance monitoring and statistics
• Connection pooling and tuning

Quick Decision Guide

Optimization Workflow

Phase 1: Diagnosis

1. Capture slow query with EXPLAIN ANALYZE
2. Identify bottleneck: sequential scan, expensive join, high cost estimate
3. Understand data volumes and selectivity

Phase 2: Root Cause Analysis

1. Check for missing indexes on filter/join columns
2. Review join strategy (nested loop vs hash vs merge)
3. Analyze cardinality estimates vs actual rows

Phase 3: Optimization

1. Add indexes (B-tree, hash, GIN, BRIN as appropriate)
2. Rewrite query to reduce complexity (CTEs, window functions)
3. Consider denormalization or materialized views
4. Measure impact: runtime, cost, resource usage

Phase 4: Validation

1. Test optimization with production data volume
2. Monitor side effects (index maintenance overhead, storage)
3. Document solution and performance improvement

Best Practices

Performance Analysis

• ✅ Establish Baseline: Measure query runtime, cost, resource usage before optimization
• ✅ Use EXPLAIN ANALYZE: Always run with ANALYZE; never guess at performance
• ✅ Understand Costs: Lower query cost estimate usually correlates with faster execution
• ✅ Check Selectivity: Filters with poor selectivity waste I/O; indexes on selective columns help most

Indexing Strategy

• ✅ Index Selective Columns: Indexes help when filtering/joining on high-cardinality columns
• ✅ Avoid Over-Indexing: Every index slows writes; add only indexes that measurably help
• ✅ Composite Indexes: Order by selectivity (most selective first) and join key order
• ✅ Partial Indexes: Index only relevant subset (e.g., WHERE active = true)

Query Optimization

• ✅ Filter Early: Push WHERE clauses down before joins
• ✅ Minimize Shuffles: Avoid sorting/aggregating large result sets; use indexes for ordering
• ✅ Use Window Functions: More efficient than self-joins for ranking, running totals
• ✅ Denormalize Strategically: Trade write complexity for read speed when appropriate

Common Pitfalls

Dependencies

• senior-data-engineer - For schema design and performance mentorship