Banking Data Mesh Migration: Starburst→Trino

Challenge

A Tier-1 bank was struggling with mixed Trino/Starburst versions across different business units, leading to rising costs, upgrade friction, and inconsistent governance across their federated data architecture.

Solution

We migrated the entire bank to a unified Trino-based data mesh while maintaining zero downtime and improving performance across all domains.

Key Components

• Federated Query Engine: Unified Trino cluster replacing mixed Starburst/Trino deployments
• Domain Isolation: Resource groups and workload isolation for each business unit
• Governance Framework: Centralized policy management with domain-specific controls
• Migration Strategy: Blue/green deployment with automated drain and cutover

Implementation

• Blue/Green Migration: Zero-downtime cutover behind existing gateway
• Drain Automation: Intelligent query routing and graceful shutdown
• Resource Optimization: Spill tuning and memory management
• MERGE Audit: Data consistency validation across all domains

Results

• 35% reduction in infrastructure costs through unified architecture
• 48% faster median query latency across all business units
• Zero downtime migration with automated rollback capabilities
• 100% compliance with banking regulations through improved governance

Technology Stack

• Query Engine: Trino with Starburst Enterprise features
• Storage: Apache Iceberg with cross-domain data sharing
• Governance: OPA-based policy engine with RBAC
• Monitoring: Custom dashboards with domain-specific metrics

Business Impact

The unified data mesh enabled:

• Consistent data access patterns across all business units
• Reduced operational overhead with centralized management
• Faster regulatory reporting through improved governance
• Cost savings that funded additional data initiatives