Cloud Migration, Any Direction

GCP to AWS, AWS to GCP, Azure to either, or out of the data centre altogether. We map every dependency, replicate the data, rehearse the rollback and move traffic in phases, so cutover is a planned event and not a long weekend.

GCP → AWS GKE, Cloud SQL, GCS and IAM onto their AWS equivalents

AWS → GCP EKS, RDS, S3 and IAM onto Google Cloud

On-Prem → Cloud VMware and bare metal out of the data centre

Azure → AWS / GCP AKS, Azure SQL and Blob Storage, re-homed

Every Path. Every Workload.

Compute, Kubernetes, databases, storage, identity, DNS and the pipelines that deploy them. A migration is only finished when all of it runs on the target and the source can be switched off.

GCP to AWS

Compute Engine to EC2 and Auto Scaling
GKE to EKS
Cloud SQL to RDS or Aurora
Cloud Storage to S3
Pub/Sub to SNS, SQS or MSK
Cloud IAM to IAM roles and IAM Identity Center
Cloud DNS to Route 53

AWS to GCP

EC2 to Compute Engine and managed instance groups
EKS to GKE
RDS and Aurora to Cloud SQL or AlloyDB
S3 to Cloud Storage
SNS and SQS to Pub/Sub
IAM to Cloud IAM and Workload Identity
Route 53 to Cloud DNS

On-Prem to Cloud

VMware and bare-metal servers to EC2 or Compute Engine
Self-managed databases to managed databases
NAS and SAN to object and block storage
Hardware load balancers to cloud load balancing
Site-to-site VPN or Direct Connect for the transition
Data centre exit plan with a decommission date

Azure to AWS or GCP

AKS to EKS or GKE
Azure SQL to RDS or Cloud SQL
Blob Storage to S3 or Cloud Storage
Entra ID kept as the identity provider through SSO
Key Vault secrets to Secrets Manager or Secret Manager

Databases and Data

Online replication with AWS DMS or Google Database Migration Service
Change data capture until the cutover minute
Row count and checksum validation
Bulk object copy with DataSync or Storage Transfer Service
Replication lag watched, not assumed

Kubernetes and Containers

Cluster to cluster moves across clouds
VM workloads re-platformed onto Kubernetes
Helm charts and GitOps with Argo CD
Velero backup and restore
Container registry moves between ECR and Artifact Registry

Infrastructure as Code

Target environment written in Terraform or OpenTofu
Separate state per environment
CI/CD pipelines repointed at the new cloud
Secrets and certificates moved, not copied by hand
Everything handed over in your repositories

Security and Residency

Least-privilege IAM from the first day
Indian regions where data has to stay in India
Audit logging on before the first workload lands
Encryption at rest with managed keys
Network segmentation carried across, not flattened

Cost

Run cost estimate before anything moves
Egress charges for the transfer itself
Right-sizing on the target, not a copy of the old sizes
Savings Plans or committed use once usage settles
Cost review a month after cutover

How a Migration Runs

Most of the work happens before cutover. The cutover itself should be the boring part.

Before Cutover

Inventory of every service, dependency, cron job and DNS record
Target architecture in code, reviewed before anything is built
Data replication running and validated for days, not hours
A rollback plan that has been rehearsed, not just written
Monitoring and alerting live on the target before traffic arrives
Run cost and transfer cost signed off

Cutover and After

Traffic shifted in phases with weighted DNS, never all at once
Source kept warm through a dual-run window
Error rates and latency compared side by side
Source decommissioned only after sign-off
Runbooks and infrastructure code handed over
A cost review once the bill has a full month in it

The checklist we work through before any cutover is written up in what has to exist before cutover.