Not: Bu vaka çalışması şu anda sadece İngilizce olarak mevcuttur. / Note: This case study is currently only available in English.
60% Cost Reduction Through Strategic Migration
How we helped Extenship save over 50% on infrastructure costs by migrating from AWS to Hetzner bare metal servers while maintaining high availability.
Client Overview
About Extenship
Industry
Technology Platform
Challenge
High cloud costs
Solution
Bare metal + Kubernetes
The Challenge
Unsustainable Cloud Costs
AWS Monthly Cost Breakdown
-
-
Cloud costs were growing faster than revenue, making the business model unsustainable. Every new feature meant higher infrastructure costs.
-
-
Heavy reliance on AWS-specific services made it difficult to optimize costs or consider alternatives without major refactoring.
-
-
Need for better GDPR compliance and data residency in Europe was becoming critical for business growth.
-
-
Paying for resources that were underutilized, but AWS pricing model made it expensive to right-size without losing performance.
Technical Implementation
Bare Metal Kubernetes Platform
1. Cost Analysis & Infrastructure Design
Hetzner Architecture
Master Node (1x)
- • AX41-NVMe (AMD Ryzen 5 3600)
- • 64GB RAM
- • 2x512GB NVMe SSD (RAID 1)
- • Cost: €150/month
Worker Nodes (2x)
- • AX41-NVMe (AMD Ryzen 5 3600)
- • 64GB RAM each
- • 2x512GB NVMe SSD (RAID 1)
- • Cost: €300/month total
Network Architecture
Private Network
Hetzner vSwitch for secure inter-node communication
Floating IPs
High availability with automatic failover
WireGuard VPN
Secure remote access for administration
Load Balancing
MetalLB for bare metal load balancing
2. Kubernetes Platform Setup
Rancher RKE2 Deployment
High Availability
- • HA control plane
- • etcd replication
- • Automatic failover
- • Load balancing
Security
- • CIS hardened
- • Network policies
- • RBAC enabled
- • Pod security
Networking
- • Cilium CNI
- • eBPF performance
- • Network policies
- • Service mesh ready
Storage Solution with Rook/Ceph
We deployed Rook/Ceph for distributed storage, providing resilient and scalable block storage for stateful workloads on bare metal infrastructure.
Ceph Configuration
- • 3 monitor nodes for quorum
- • 2 manager instances
- • SSL-enabled dashboard
- • Dedicated NVMe storage devices
Resource Allocation
- • 1TB NVMe per storage node
- • 2-4 CPU cores per OSD
- • 4-8GB memory per OSD
- • Optimized for performance
3. AWS Service Replacements
Database Migration
RDS → CloudNativePG
- • PostgreSQL operator on Kubernetes
- • Automated backups to MinIO
- • Point-in-time recovery
- • Connection pooling with PgBouncer
ElastiCache → Redis Operator
- • Master-slave replication
- • Automated failover
- • Persistence configuration
- • Sentinel for HA
Service Replacements
S3 → MinIO
- • S3-compatible API
- • Distributed deployment
- • Erasure coding
- • Lifecycle policies
ALB → MetalLB + Nginx
- • Bare metal load balancing
- • Nginx ingress controller
- • Let's Encrypt SSL
- • Rate limiting & WAF
PostgreSQL High Availability Setup
We implemented CloudNativePG for PostgreSQL high availability, providing automatic failover, backup management, and monitoring integration.
Database Configuration
- • 3-node cluster for HA
- • 200 max connections
- • Optimized memory settings
- • 100GB storage per instance
- • Ceph block storage integration
Backup Strategy
- • 30-day retention policy
- • S3-compatible backup target
- • WAL archiving (7 days)
- • Parallel backup streams
- • Point-in-time recovery
4. Self-Hosted Monitoring Stack
Metrics
- • Prometheus for collection
- • Grafana dashboards
- • Custom alerts
- • Long-term storage
Logs
- • Loki for aggregation
- • Promtail collectors
- • LogQL queries
- • Retention policies
Alerts
- • Alertmanager setup
- • Slack integration
- • On-call rotation
- • Escalation policies
5. Migration Execution
Phase 1: Infrastructure Setup
Hetzner servers provisioning and Kubernetes installation
- • Bare metal server setup
- • Network configuration
- • Kubernetes deployment
- • Storage setup with Ceph
Phase 2: Service Deployment
Deploy open-source alternatives and test functionality
- • PostgreSQL and Redis setup
- • MinIO object storage
- • Application deployment
- • Load balancer configuration
Phase 3: Data Migration
Sync data from AWS to Hetzner infrastructure
- • Database replication setup
- • Object storage sync
- • Application state migration
- • Data validation
Phase 4: Traffic Cutover
Gradual traffic migration and AWS shutdown
- • DNS preparation
- • Traffic splitting
- • Monitoring and validation
- • AWS resource cleanup
6. Backup & Disaster Recovery
Backup Strategy
- Velero for Kubernetes backup
- Automated database backups to MinIO
- Off-site backup to Backblaze B2
- 30-day retention policy
Disaster Recovery
- Documented recovery procedures
- Regular DR drills
- RTO: 4 hours, RPO: 1 hour
- Infrastructure as Code for quick rebuild
Results Achieved
Dramatic Cost Reduction & Performance Gains
Monthly Cost Comparison
Previous AWS Costs
Current Hetzner Costs
Performance Benefits
Operational Benefits
"We worked with Fizyonops on alternative infrastructure models to avoid high cloud costs. Thanks to the system they built on our own hardware, we saved over 50% of our infrastructure budget."
Emre Tekin
Founder, Extenship
Challenges Overcome
Building Managed Services on Bare Metal
Technical Challenges
Automated Failover
Implemented Keepalived with floating IPs for automatic failover without managed load balancers.
Managed Service Alternatives
Built comparable functionality using open-source operators and careful configuration.
High Availability
Achieved through careful architecture design and redundancy at every layer.
Solutions Implemented
Automation Everything
Ansible playbooks and GitOps for consistent deployments and configuration management.
Monitoring & Alerting
Comprehensive monitoring to catch issues before they impact availability.
Regular Maintenance
Automated updates and maintenance windows to keep systems secure and performant.
Key Takeaways
Lessons from Cloud to Bare Metal Migration
Cloud Isn't Always Cheaper
For predictable workloads, bare metal can offer significant cost savings while providing better performance and full control.
Open Source Maturity
Modern open-source tools can provide enterprise-grade functionality comparable to managed cloud services.
Automation is Critical
Comprehensive automation and monitoring are essential to achieve cloud-like reliability on bare metal infrastructure.
Want to Reduce Your Infrastructure Costs?
Let's analyze your current infrastructure and explore cost optimization opportunities.