Best Practices

Performance and cost factors

• How many bytes are read from query
• Shuffling: how many bytes does your query pass to each slot
• Computation: CPU times
• Outputs: how many bytes are written

Reducing bytes read

• Avoid SELECT *
• LIMIT does not limit data scanned. Use "Maximum bytes billed" in query settings instead
• Use partitioning or clustering
• Use denormalization
• Use Materialized Views to keep aggregated data
• Use cached results
• Avoid external data source

SQL best practices

• Avoid self joins use Window function instead
• Avoid cross joins, pre aggregate data or use Window function
• Avoid DML single row modifications