What is a cloud exit?
Moving production workloads from cloud providers (AWS, GCP, Azure) to dedicated servers or bare-metal infrastructure that you control. Also called cloud repatriation. You stop renting compute and start owning it.
Is bare-metal actually reliable enough for production?
Yes. The internet ran on bare-metal for over 20 years before "the cloud" became a thing. Netflix, Facebook, and Google all run on their own hardware. Reliability comes from redundancy and engineering — not from a cloud provider's brand name. We deploy 3-node clusters with automatic failover. A single server failure causes zero downtime.
How long does a migration take?
8–12 weeks from assessment to production, depending on complexity. A simple setup (one database, one app) can be faster. A complex environment (multiple databases, messaging, Kubernetes) takes the full 12 weeks. Every step has a rollback plan.
Will I need a bigger ops team?
No. This is the central myth of cloud marketing — that it's so much easier you barely need anyone to operate it. We've never seen that be true. The team you have for cloud is the team you need for bare-metal. 90% of the expertise is identical. The other 10% is stuff like "how to SSH into a server" which your team already knows.
What about auto-scaling?
Hardware is so cheap that you can massively over-provision and still save 70% compared to cloud. A dedicated server with 96 cores and 256GB RAM handles traffic spikes that would require auto-scaling in the cloud. Buy the baseline, rent the spike — but most companies never actually need the spike.
Can I keep some workloads in the cloud?
Absolutely. Hybrid is fine. Move what saves money (databases, compute, storage), keep what doesn't make sense to move (deeply integrated serverless functions, ML training pipelines). We'll tell you which workloads should stay.
What if the migration fails?
Every step has a rollback plan. Your cloud infrastructure stays running until you explicitly confirm the cutover. We don't touch your source systems. If anything goes wrong, you're back on cloud in minutes. We've never had a migration fail, but the rollback plan is always ready.
What databases do you support?
PostgreSQL, MongoDB, Cassandra, Elasticsearch, and Redis. We deploy them with production-grade HA: Patroni for PostgreSQL, replica sets for MongoDB, gossip protocol for Cassandra, cluster mode for Elasticsearch, Sentinel for Redis. If it's open source and you're running it in the cloud, we can deploy it on bare-metal.
Can Kubernetes run on bare-metal?
Yes. K3s is CNCF-certified Kubernetes. It runs identically to EKS/GKE. We add MetalLB for LoadBalancer services, cert-manager for TLS certificates, and Velero for cluster backups. Your Helm charts and manifests work unchanged.
How do you handle disaster recovery?
Every service has automated backups to S3-compatible storage. PostgreSQL has point-in-time recovery (restore to any second). MongoDB has continuous oplog capture. Cassandra has daily snapshots. Kubernetes has Velero. We test restores — because backups that haven't been tested aren't backups.
What happens if a server fails?
Every service is clustered across 3+ nodes. Database clusters auto-failover in 10–30 seconds via Patroni (PostgreSQL) or replica election (MongoDB). HAProxy detects and reroutes in under 5 seconds. Cassandra continues serving from remaining nodes. You lose a server, not a service.
How is monitoring handled?
Prometheus for metrics, Grafana for dashboards, Loki for logs, Alertmanager for alerts. Every server has node_exporter. Every database has its own exporter. Pre-built dashboards for every service. Alerts for disk, replication lag, node health, backups, certificates. Multi-channel: email, Slack, PagerDuty.
How is security handled?
Hardened bastion host as the only SSH entry point. fail2ban for brute-force prevention. PSAD for port scan detection. UFW firewall on every server (default-deny). Private VLAN for all internal communication. Encryption at rest for MongoDB. TLS automation for certificates. Audit logging via auditd. No database port exposed to the internet — ever.
Do you support any hosting provider?
Any provider where we get SSH access to dedicated servers. Hetzner, OVH, Equinix, Vultr, Leaseweb, your own colo, private cloud — we're agnostic. We don't resell servers. You have the direct relationship with your provider.
How do you calculate savings?
Your verifiable cloud bill (from your AWS/GCP/Azure billing dashboard) minus the new bare-metal infrastructure cost (server rental + bandwidth from your hosting provider). We use real documents, not estimates. Both numbers are independently verifiable.
What's the minimum cloud spend?
$20,000/month. Below that, the absolute dollar savings may not justify the migration effort and our management fee. We'll be honest about this during assessment — if the math doesn't work, we'll tell you to stay.
What if savings are less than expected?
Our fee is a percentage of actual savings. Less savings = lower fee. Zero savings = zero fee. The risk is entirely on our side. We're incentivized to maximize your savings because that maximizes our fee.
Can I switch tiers later?
Yes. Upgrade or downgrade with 30 days notice. Most clients start with Silver and upgrade to Gold once they see the value of proactive management.
What happens if I want to cancel?
90-day notice. We hand over everything: Ansible playbooks, documentation, runbooks, monitoring dashboards, credentials, backup configurations. It's your infrastructure — we built it, you own it. No hostage-taking.
Who owns the data and infrastructure?
You do. Always. The servers are in your name with your hosting provider. The databases contain your data. We manage it — you own it. If we part ways, nothing changes except who's watching the dashboards.
Is there a minimum contract?
12 months for managed operations. This ensures we have time to fully stabilize the infrastructure and you have time to realize the savings. Consulting engagements (assessment, architecture review) are one-time fixed fee with no ongoing commitment.
How are you different from traditional IT services companies?
The traditional IT services model scales by adding headcount. More servers to manage? Hire more engineers. More complexity? Add more people. The client pays per person, per hour — regardless of outcomes. The incentive is to keep teams large and engagements long.
We're the opposite. Our model is built on outcomes, not hours. We charge a percentage of the cost savings we deliver — not for the number of people on the project. If we don't save you money, you don't pay.
This is possible because of how we work. Instead of staffing dozens of junior engineers, our deployments are executed by a small team of senior infrastructure architects who leverage AI-powered automation for the heavy lifting — configuration generation, deployment orchestration, validation, and monitoring setup. What traditional firms staff a 15-person team for over 6 months, we deliver with a handful of experts in 8–12 weeks.
The result: you get deeper expertise, faster delivery, lower cost, and a pricing model where our success is directly tied to yours.
Do you offer fixed-cost engagements?
Yes, absolutely. If your procurement process or internal policy requires a fixed-cost contract, we're happy to structure it that way.
In practice, the number we arrive at is almost always derived from the same formula — a percentage of your projected annual savings. So whether you're paying us monthly as a percentage or annually as a fixed number, the underlying economics are the same. You're paying a fraction of what you save, and we're incentivized to maximize that savings.
The model works either way — as it should. We're flexible on structure, not on alignment.