AI-powered performance recommendations for Amazon Redshift

This article demonstrates how to build an AI-powered solution using Amazon Bedrock and Lambda to automatically analyze Amazon Redshift performance telemetry and deliver prioritized optimization recommendations.

• Collector Lambda runs diagnostic SQL queries and CloudWatch metrics, pre-computes performance signals, and writes telemetry to S3
• Analyzer Lambda reads telemetry, builds structured prompts with signal-to-metric correlations, and calls Amazon Bedrock Claude for recommendations
• Solution detects table-level issues like row skew, stale statistics, and sub-optimal keys, plus runtime problems like disk spill and small-insert bursts
• Recommendations include priority levels, categories, signal sources, explanations, SQL actions, and expected impact estimates
• EventBridge triggers daily 24-hour runs; SNS sends email summaries of top 10 recommendations to subscribers
• Signal-based design produces specific, workload-targeted advice rather than generic best-practice guidance

The solution enables automated daily performance reviews for Amazon Redshift Serverless with minimal operational overhead and cost-effective Amazon Bedrock usage.