Skip to content

Cloud migration: a parallel environment, a short cut-over, a way back

We move systems into AWS, out of AWS to cheaper providers, and from standalone servers into Kubernetes. The new environment is built next to the old one, data is replicated while the old system keeps working, and traffic switches in an agreed window. The old environment stays until you confirm the new one.

When you need this

  • 1

    The AWS bill no longer fits the business and the same load would cost less elsewhere

  • 2

    You have outgrown a hosting provider and need AWS: managed databases, more regions, EKS

  • 3

    Servers set up by hand years ago need to move into a cluster

  • 4

    The product has to be hosted in another country or region

  • 5

    The migration keeps being postponed because nobody wants to risk downtime

What's included

  • Inventory

    Services, databases, files, cron jobs, DNS records, certificates, external integrations and IP allowlists. We also find what depends on provider-specific services and what can be dropped.

  • Target design and cost

    What the system looks like on the new side. On AWS that means EKS or EC2, RDS, S3 and the VPC layout; when leaving AWS, what replaces each managed service. You see the expected bill before you decide.

  • Parallel environment

    The new infrastructure is created with Terraform next to the old one. Services are deployed there from the same repository and checked without touching production.

  • Data sync

    Databases move by replication from the old side to the new one, so the copy stays current until the cut-over. Object storage is synced to or from S3 with rclone in several passes. Large volumes are copied in advance.

  • Cut-over

    The DNS TTL is lowered beforehand. In the window, writes stop on the old side, replication catches up, the new database is promoted and traffic is switched. Downtime is the length of this step.

  • Rollback path

    The old environment stays running and untouched until you sign off. The conditions for switching back are written down before the window, not during it.

The bill follows the load

A cost review ends with a report like this: what you paid before, what you pay now, and where the difference came from.

billing · cost reviewmonthly, USD

Monthly cost

Before
$10,400
After
$6,200
Δ per month
−40%

Cost by month

M1M2M3M4M5M6
BeforeAfter

Cost by service

ServiceBeforeAfterΔ
EC2 / EKS compute$5,200$2,900−44%
RDS$2,100$1,500−29%
NAT + data transfer$1,600$650−59%
S3 + EBS$900$550−39%
Other$600$6000%

Example of the report format with illustrative figures. The real saving depends on the account and is shown to you before anything is changed.

How the work goes

  1. 1

    Call and access

    You tell us what moves and why. We get read access to the current environment and to the bill.

  2. 2

    Plan and price

    Inventory, target architecture, cost comparison, migration order and expected downtime. The price is agreed before work starts.

  3. 3

    Build and rehearse

    The new environment goes up, data is synced, and we do a trial cut-over on a copy. Load and a restore from backup are tested on the new side.

  4. 4

    Cut-over and sign-off

    Traffic switches in the agreed window and we watch the system. After your sign-off the old environment is shut down.

Questions

  • Terraform
  • Kubernetes
  • AWS EKS
  • AWS EC2
  • AWS RDS
  • AWS S3
  • Helm
  • Hetzner
  • Cloudflare
  • PostgreSQL
  • MySQL
  • rclone

Describe the task — get a plan and a price

Write what you have and what you need. We answer within one business day with questions or a concrete proposal.