How Amazon Achieved Full Stack Observability Across 400 Offices with Amazon OpenSearch Serverless

This article describes how Amazon Corporate Infrastructure Services built a Full Stack Observability platform using Amazon OpenSearch Serverless to monitor 330,000+ employees across 400 offices globally.

• Consolidated fragmented monitoring from domain-specific tools into unified platform eliminating data silos
• Achieved 5-minute Mean Time to Detect (MTTD) target, down from 30-60 minutes baseline
• Used Amazon OpenSearch Ingestion for data normalization, transformation, and enrichment across multiple telemetry sources
• Implemented three-layer architecture: data sources, processing/storage, and insights/alerting with OpenSearch Dashboards
• Scaled from 3 pilot sites to 24 sites in Phase 1 using repeatable ingestion patterns and Infrastructure as Code
• Achieved 99.9% platform uptime and 220% projected annual ROI through engineering time and productivity savings
• Key lessons: start with business outcomes, embrace open standards (OpenTelemetry), design for scale, invest in data quality, and enable self-service dashboards

The serverless approach eliminated operational complexity while enabling rapid expansion and proactive issue detection across Amazon's global infrastructure.