Astuto Achieves 15 Daily Deployments with Zero-Downtime AWS to GCP Migration

How Astuto leveraged Facets Platform to migrate to GCP while maintaining AWS dependencies, achieving multi-cloud operations without application changes.

Strategic Value Delivered by Facets

Before

• AWS-native ECS-based Architecture: 30+ services deeply integrated with AWS services, making migration challenging

• Configuration Management Complexity: Critical systems where teams feared making changes that could break production

• Multi-Cloud adoption: Unable to adopt multi-cloud strategy without extensive application rewrites and months of migration effort

After

• Seamless Multi-Cloud Operations: Compute on GCP and data services on AWS, connected via workload identity

• 15 Daily Deployments: Across environments with change failure rates below 1%

• Zero-Application-Change Migration: Simple configuration toggles to switch between clouds

• Single-Click Environment Deployment: Capability for customer premises and new regions across any cloud

The Challenge: Complex AWS-Native Migration

Technology Transformation Complexity

Astuto needed to migrate from Amazon ECS to Kubernetes while expanding to new regions and customer environments. With 30+ microservices and complex Apache Airflow (DAGster) orchestration, the migration seemed daunting.

Configuration Management Crisis

The DAGster configuration had difficult to port. Teams were afraid to modify configs due to hardcoded environment values creating a fragile system where any change could cascade into production failures.

Multi-Cloud Strategy Challenge

Astuto's architecture was initially designed to be fully AWS native, with many services having no GCP equivalent. The platform relied heavily on:

• 23 DynamoDB tables (no direct GCP equivalent)
• 9 SQS queues (no direct GCP equivalent)
• Lambda functions
• Step Functions, EventBridge, and Cognito

Migrating these AWS-native services would require significant developer time to rewrite applications for GCP equivalents, and the risk of breaking production systems was high. Traditional migration approaches estimated months of effort and potential service disruptions.

How Facets Transformed Astuto's Platform

Phase 1: ECS to Kubernetes Transformation

• Automated Discovery: Scripts discovered all services and dependencies, converting ECS to Kubernetes
• DAGster Simplification: Dereferenced configurations with logical references, making changes safe
• Kubernetes-Ready Architecture: Teams focused on applications while Facets handled the container orchestration complexity

Phase 2: Multi-Cloud Blueprint Architecture

The breakthrough came with Facets' multi-cloud blueprint approach, which accelerated the migration by allowing Astuto to:

• Move services with GCP equivalents immediately: Compute workloads, Kubernetes services, and containerized applications migrated seamlessly to GKE
• Keep AWS-native services running: DynamoDB, SQS, Lambda, and other AWS-specific services continued operating without modification
• Plan future migrations incrementally: AWS-native services marked for future migration when business priorities align

This hybrid approach eliminated the need for risky rewrites while enabling immediate multi-cloud benefits:

• Configuration-Based Switching: Services switch between AWS/GCP with simple flavor changes
• Cross-Cloud Authentication: OIDC connected GCP workload identities to AWS IAM roles
• Intelligent Placement: Compute and portable services moved to GKE on GCP, AWS-native services remained operational

Phase 3: Zero-Downtime Migration to GKE

• Created parallel GKE environment on GCP alongside AWS for QA and production
• Service-by-service validation and migration to GKE clusters
• Simple DNS updates for traffic switching
• Instant rollback capability maintained between clouds

Key Results

• 15 deployments/day with less than 1% change failure rate
• Configuration changes: From being unpredictable to predictable
• New region deployment: From weeks to days
• Spot instances enabled for all lower environments
• Complete developer self-service achieved

Infrastructure Scale

• 167 service resources managed through single platform
• 30+ microservices migrated to GKE on GCP
• 23 DynamoDB tables accessed seamlessly from GKE pods

Business Expansion Capabilities

• Single-click multi-cloud environments: Launch complete environments for customer premise deployments
• Rapid regional expansion: Deploy to new regions across any cloud provider with configuration changes
• Customer-specific deployments: Spin up isolated environments for enterprise customers on their preferred cloud

What Made the Difference?

Blueprint-First Approach: Reusable blueprints that could target any cloud, enabling testing in lower environments and cloud switching without application changes.

Phased Migration Strategy: Move what's portable now, keep AWS-native services for future migration, eliminating the all-or-nothing risk.

Configuration as Code: Dereferenced environment values transformed infrastructure management with centralized configs and safe testing.

Incremental Migration: Service-by-service migration at Astuto's pace with continued feature development and instant rollback capabilities.

What's Next

With the multi-cloud foundation established, Astuto now has the ability to:

• Expand to new regions instantly across AWS, GCP, or Azure based on customer requirements
• Deploy isolated environments for enterprise customers on their preferred cloud provider

This multi-cloud capability positions Astuto to rapidly scale their OneLens platform globally and meet enterprise customer requirements for data residency and cloud preferences.