A business application needs frequent small transactional updates and constraints on operational records. A reporting workload needs large aggregations across historical data. Which high-level division is most appropriate?

1. Run every historical aggregation as a full scan on the busiest transactional instance without assessing impact.
2. Treat Redshift as interchangeable with PostgreSQL for every high-concurrency row transaction without workload evaluation.
3. Use the reporting output CSV as the only transactional record system.
4. Use an appropriate RDS or Aurora transactional database for the application and Redshift for the analytical warehouse, connected by a suitable data pipeline.