← Back to projects

MAINROUTE Cloud Migrations

Enterprise Infrastructure Transformation

Chief Technology OfficerDec 2015 - Sep 2025

Overview

As CTO at MAINROUTE, I led technical strategy and infrastructure modernization for enterprise clients across multiple Brazilian states. Over 10 years, I directed five major cloud migration projects, ranging from on-premises VMware to AWS/Azure, cloud-to-cloud transitions, and critical ERP database conversions - all with minimal downtime and measurable cost savings.

The Challenge: Enterprise Migrations at Scale

Enterprise infrastructure migrations are high-risk, high-impact projects that can't fail:

  • Zero tolerance for data loss: Financial and ERP systems require perfect data integrity
  • Minimal downtime requirements: 24/7 operations mean migration windows measured in hours, not days
  • Legacy system complexity: 10+ year old applications with undocumented dependencies
  • Compliance requirements: LGPD (Brazilian GDPR) and financial industry regulations
  • Cost constraints: Must demonstrate ROI within 18-24 months

Migration Projects Delivered

Project 1: Portocred S/A - SQL Server Migration to Azure

Context: Financial institution with 500GB SQL Server database running on aging on-prem hardware

Approach:

  • Assessed current SQL Server environment: identified missing indexes, deprecated configurations, performance bottlenecks
  • Designed SQL Server AlwaysOn Availability Group architecture for high availability
  • Chose Azure SQL Managed Instance over VM-based SQL Server for reduced operational overhead
  • Implemented Database Migration Service (DMS) for minimal-downtime cutover
  • Created rollback procedures and tested failover scenarios

Result: 99.95% uptime achieved. 40% reduction in database hosting costs. 24x7 DBA support delivered with documented failover procedures.

Project 2: Retail Client - On-Prem to AWS Migration

Context: Multi-location retail operation with VMware infrastructure and MySQL databases

Technical Decisions:

  • MySQL Strategy: AWS RDS Multi-AZ instead of self-managed EC2 - trade automated backups/patching for slightly higher cost
  • Application Servers: Kubernetes (EKS) for containerized workloads to enable auto-scaling
  • Network Design: VPN + Direct Connect hybrid for gradual migration without cutting over all systems at once
  • IaC: Terraform modules for reproducible infrastructure across dev, staging, production

Performance Optimization:

  • Identified I/O bottlenecks through CloudWatch and MySQL slow query logs
  • Recommended RDS instance type changes (memory-optimized instances for better buffer pool hit ratio)
  • Implemented ElastiCache (Redis) for session storage to reduce database load

Result: Peak-load database stability improved by 60%. Auto-scaling enabled handling of Black Friday traffic without over-provisioning year-round.

Project 3: Cloud-to-Cloud Migration (Azure → AWS)

Context: Client needed to consolidate from Azure to AWS due to vendor standardization

Challenges:

  • Azure-specific services (App Service, Cosmos DB) had no 1:1 AWS equivalents
  • Active-active architecture required both clouds operational during transition
  • DNS cutover had to be instantaneous to avoid downtime

Solution:

  • Mapped Azure services to AWS equivalents: App Service → ECS Fargate, Cosmos DB → DynamoDB
  • Implemented database replication (Azure Cosmos → DynamoDB) using custom Python sync scripts
  • Blue-green deployment strategy: ran both clouds in parallel, gradual traffic shift via Route 53 weighted routing

Result: Zero-downtime migration. Decommissioned Azure resources after 30-day validation period.

Key Technical Decisions & Trade-offs

Managed Services vs Self-Managed

Decision Framework:

  • Choose Managed (RDS, Managed Instance): When operational overhead (patching, backups, HA) outweighs cost difference
  • Choose Self-Managed (EC2 + PostgreSQL): When specialized configurations (custom extensions, kernel tuning) are required
  • Result: 70% of migrations used managed services - freed client IT teams to focus on application development instead of database operations

Migration Strategy: Big Bang vs Gradual

Lessons Learned:

  • Big Bang (single cutover weekend): Only for smaller systems (< 100GB databases, < 5 applications). Higher risk but faster ROI.
  • Gradual (hybrid cloud period): Essential for mission-critical systems. More expensive (running two clouds), but de-risks migrations.
  • Implemented: Hybrid approach 80% of the time - non-critical workloads first, then databases, then frontend applications.

DevOps & Automation

  • Infrastructure as Code: Terraform for AWS/Azure resource provisioning. Version-controlled infrastructure changes in Git for audit trail and rollback capability.
  • CI/CD Pipelines: GitHub Actions + Jenkins for automated testing and deployment. Database migration scripts tested in staging before production cutover.
  • Automation Scripts: Python scripts for database performance assessments, resource right-sizing recommendations, and cost analysis. Reduced manual assessment time from weeks to days.
  • Monitoring: Zabbix + CloudWatch for unified monitoring across on-prem and cloud. Telegram alerts for infrastructure issues with automated ticket creation in service desk.

Cost Optimization Achievements

35-50%
Average infrastructure cost reduction over 6-month period
5
Major migrations delivered with minimal downtime
99.9%
Availability maintained during migration windows
Zero
Data loss incidents across all migration projects

Technical Stack Across Projects

AWSAzureVMware ESXiTerraformKubernetesDockerSQL ServerMySQLPostgreSQLPythonPowerShellZabbixCloudWatchGitHub Actions

Risk Management & Governance

  • Pre-Migration Assessments: Comprehensive discovery of all dependencies, database sizes, network bandwidth requirements. Documented in technical design specifications.
  • Rollback Plans: Every migration had tested rollback procedures. Database snapshots taken immediately before cutover. DNS TTLs reduced to 60s for fast reversion.
  • Stakeholder Communication: Weekly status reports to C-level executives. Migration runbooks shared with client IT teams. Post-migration knowledge transfer sessions.
  • Compliance: LGPD compliance validated through data residency requirements (Brazil region). Security group rules, encryption at rest, and audit logging implemented per financial industry standards.

Lessons Learned

  • Discovery Phase is Critical: 90% of migration issues stem from incomplete discovery. Invest 2-3 weeks mapping all dependencies - saves months of troubleshooting later.
  • Test Rollback Procedures: Don't assume backups work. Actually test restoring from backup in a non-production environment before migration day.
  • Gradual Beats Big Bang: Hybrid cloud periods are expensive but worth it. Can validate one workload at a time instead of betting entire business on a single cutover weekend.
  • Cost Optimization is Ongoing: Right-sizing analysis every quarter revealed 15-20% additional savings after initial migration. Overprovisioning is common early on.
← Back to all projects