Services
/
04 — Cloud & DevOps
Project + managed

Migrations that finish on the weekend they were planned for.

Assessment, migration and day-two operations across AWS, Azure and Google Cloud — with all infrastructure defined in code and a rehearsed failover before anything cuts over.

Engagement
9–27 weeks
Model
Project + managed
Day two
Managed support
Chapter one
Overview

Cloud projects rarely fail technically

They fail on the third party nobody asked about. A fixed IP allowlist controlled by a partner who needs months to change it. A licence tied to a MAC address. A batch job whose owner left the company years ago. These are what turn a short migration into a long one, and every one of them is discoverable in the first week if somebody looks.

So our assessment is deliberately dull and exhaustive. By the time we quote you a cutover weekend, the sequence has been rehearsed against production data and every external dependency has a name and a date beside it.

Suited to
Companies facing a hardware refresh quoted higher than several years of cloud spend
Teams with a single data centre and no tested disaster recovery
Businesses whose releases require an evening and most of the team
Organisations with data residency obligations in India, the EU or the GCC
Chapter two
Scope of work

What the service covers, usually in this order

01
Cloud migration

Assessment, dependency mapping, a lift-or-refactor decision per workload, and a rehearsed cutover with a rollback path.

02
Cloud infrastructure

Networking, identity, backup and disaster recovery defined in Terraform and committed to your repositories.

03
DevOps automation

Environment provisioning, secret management and observability, so a new environment is a pipeline run rather than a project.

04
CI/CD

Build, test and deploy pipelines that make releases uneventful, including the rollback that usually goes untested until the day it is needed.

Chapter three
Outcomes

What changes operationally

01
Releases stop being events

Deployments move into working hours, because rollback is tested and the pipeline checks what people used to check by hand.

02
Recovery is proven rather than assumed

Failover is rehearsed on a schedule and the result is written down, so your recovery time is a measurement rather than an estimate.

03
Run cost becomes visible

Tagging, budgets and a monthly review. Idle and oversized resources are identified and reported rather than quietly renewed.

04
You could rebuild it without us

Everything is in Terraform in your repositories. That is the standard we hold ourselves to, not a claim about how helpful we are.

Chapter 3b
Why migrations slip

What the assessment is looking for

Every row below has delayed a real migration. All of them surface in the first week if someone goes looking, which is the entire purpose of the assessment.

The blocker
How we handle it

A fixed IP allowlist controlled by a partner who needs months to change it

A third-party register built in week one, with migration waves sequenced around their calendar

A licence tied to hardware, discovered during cutover

A licence audit completed before any target architecture is priced

A batch job whose owner left the company years ago

Dependency mapping from observed network traffic rather than from documentation

Disaster recovery that has never actually been tested

A full failover rehearsal on production data before the cutover window opens

Chapter 3c
After go-live

The work that starts after the migration ends

Monitoring and alerting

Dashboards you can read, with alerts routed to named engineers.

On a rota
Patch window

A scheduled window with a rollback path, agreed with you in advance.

Monthly
Cost review

Tagging and budgets reviewed, with idle and oversized resources reported.

Monthly
Recovery test

Failover exercised and the real recovery time written down each quarter.

Quarterly

All infrastructure is defined in Terraform and held in your repositories, so this arrangement is one you can end without a rebuild.

Chapter four
Delivery process

Assess exhaustively, rehearse, then cut over

I
Discover

Workload inventory, dependency mapping, licence and third-party audit, and a target architecture with costs attached.

1–2 weeks
II
Plan

Landing zone design, identity and network model, migration waves, and a rollback plan for each wave.

1–2 weeks
III
Build

Infrastructure as code, pipeline build, non-production migration, then application migration wave by wave.

6–20 weeks
IV
Launch

Full failover rehearsal on production data, then the planned cutover window with the rollback path live.

1–3 weeks
V
Operate

Managed support, monitoring, patch windows, cost review, and a quarterly disaster recovery test.

Ongoing
Chapter five
Tools and platforms

Certified on all three, incentivised by none

We hold no reseller margin that depends on which cloud you choose. For regulated Indian workloads we also deploy to private cloud and on-premise Kubernetes.

Cloud
AWS · Azure · Google Cloud · private cloud
Infrastructure
Terraform · Kubernetes · Docker
Pipelines
GitHub Actions · GitLab CI · Azure DevOps
Observability
Grafana · Prometheus · CloudWatch · Loki

We avoid: provider-specific services that make leaving expensive; Kubernetes on projects that do not need it

See the full technology page →
Chapter six
In practice
[ architecture diagram ]
Cloud & DevOps
Infrastructure

Legacy systems moved off ageing on-premise hardware

Infrastructure that needed constant maintenance, could not scale, and was costing more each year to keep running. We migrated the business applications to cloud, automated the deployment process, rebuilt monitoring, and put secure backup and disaster recovery in place.

−34%
infrastructure cost
0 hours
unplanned downtime
99.9%
availability
Practice: Cloud Solutions · Duration: 9 weeks · Team: Cloud consultant, DevOps specialist · Technology: AWS, Docker, Terraform
Read the write-up →
Before the
cutover

Usually the one your operations team can already support, or the one your enterprise agreement already covers. Where there is no such constraint we default to AWS, or to Azure for Microsoft-heavy estates. We hold no reseller margin on any of them, so we have no commercial reason to push you either way.

Every wave has a rollback path that is built and tested before the wave runs, and the failover rehearsal happens on production data before the cutover window opens. If the criteria we agreed are not met inside the window, we roll back and reschedule. That decision is made against a written checklist rather than in the moment at two in the morning.

Yes, and most clients do, but it is a separate agreement rather than an assumption. Monitoring, patch windows, cost review and a quarterly recovery test. Because everything is defined in Terraform and held in your repositories, moving to your own team or another provider later is a decision rather than a migration.

Workloads are deployed in the region your obligations require, including backups and log retention, which is where residency requirements are most often missed. Where the requirement rules out public cloud entirely, we deploy to private cloud or on-premise Kubernetes instead.

Find out what a migration would actually cost

We run assessments as standalone paid work: current state, target architecture, and a real cost comparison against what you spend now. There is no obligation to use us for the migration, and the assessment is worth having even if you decide not to move.