Skip to main content
DevOps // Cloud

The cloud is just someone else's computer. Let's configure it properly.

Landing zones, networking, identity, and cost controls designed and codified in Terraform, for a single app or a whole platform, in public cloud, on bare metal, or both.

~/anvexis/devops/cloud-architecture

$ anvexis tldr

Codify it, tag it, budget it, and never give the default VPC a production database.

$ anvexis facts

Clouds
AWS · GCP · Azure · bare metal
IaC
Terraform / OpenTofu
FinOps
Tagging · budgets · rightsizing

What a solid foundation includes

Identity & access

SSO, roles, and least-privilege policies for people and services.

Networking

Private subnets, controlled egress, and edge protection.

Cost control

Tags, budgets, alerts, and rightsizing reviews.

Everything in Terraform

Reviewable, repeatable environments for dev, staging, and production.

Cloud bills grow quietly

Idle environments, oversized databases, and forgotten snapshots add up. A monthly cost review is cheaper than the surprise.

Why cloud architecture matters

Moving to the cloud is easy; running well there is not. Without a plan, cloud accounts grow into a tangle of hand-made servers, open ports, forgotten test environments, and bills that rise every month without an obvious reason.

Good cloud architecture starts with a foundation: separate accounts or projects for each environment, private networking, central identity and access management, logging, and budgets with alerts. On top of that foundation, each workload gets the right services for its needs, whether managed databases, containers, serverless functions, or plain virtual machines.

We work across AWS, Google Cloud, and Azure, and with on-premise or hybrid setups when data residency, latency, or cost make that the better choice. Everything is built as code, so it can be reviewed, repeated, and recovered.

Choosing where things run

WorkloadUsually a good fitWhy
Odoo and business appsManaged containers or VMs plus managed PostgreSQLPredictable load, simple operations
APIs with spiky trafficContainers with autoscaling or serverlessPay for use, scale quickly
Batch jobs and AI pipelinesScheduled containers or managed batch servicesRun only when needed
Static sites and assetsObject storage plus CDNFast globally, very low cost
Regulated or local dataSpecific regions or on-premiseResidency and compliance

Foundation we build

Account structure

Separate environments with clear boundaries and billing.

Identity and access

Single sign-on, roles instead of shared keys, and least-privilege policies.

Networking

Private subnets, controlled egress, and secure access for administrators.

Logging and audit

Central logs and audit trails retained according to policy.

Cost controls

Budgets, alerts, tagging, and regular rightsizing reviews.

Infrastructure as code

Terraform modules for every resource, reviewed like application code.

How an engagement runs

  1. Stage 01

    Review

    Week 1

    Inventory current resources, costs, risks, and requirements.

  2. Stage 02

    Design

    Week 1–2

    Agree the target architecture, regions, and services.

  3. Stage 03

    Foundation

    Week 2–3

    Build accounts, identity, networking, and logging as code.

  4. Stage 04

    Migrate

    Week 3+

    Move workloads in planned waves with rollback plans.

  5. Stage 05

    Optimize

    Ongoing

    Review costs and performance after real usage.

Quick cloud health check

  • Every resource is created through code
  • No long-lived access keys for people
  • Production is separated from development
  • Budgets and alerts exist for every account
  • Backups are stored in a separate account or region
  • Unused resources are reviewed monthly

Common pitfalls we help you avoid

The first pitfall is building by hand in the console. It is quick on day one and impossible to reproduce on day ninety, so every resource we create is defined in code from the start.

The second is a single account for everything. Mixing production and testing in one account makes mistakes more damaging and costs harder to understand, so environments get clear boundaries.

The third is letting costs drift. Idle test servers, oversized instances, and forgotten storage add up quietly, which is why budgets, tagging, and monthly reviews are part of the foundation.

What we measure

MeasureWhy it matters
Monthly cost by environmentWhere the money goes and how it trends
Resources managed as codeHow reproducible and reviewable the setup is
Security findingsOpen issues from configuration scans
AvailabilityUptime of critical services against targets

Who does what

We handle

  • Architecture and landing zone
  • Infrastructure as code and migrations
  • Cost and security reviews

Your team handles

  • Business requirements and priorities
  • Cloud account ownership and billing
  • Approving changes to production

A worked example

Consider a business that grew into its cloud account over several years: production and test servers side by side, shared admin keys, and a monthly bill nobody can explain. A new landing zone separates environments, replaces shared keys with single sign-on roles, and brings resources under Terraform.

Workloads move over in waves, each with a rollback plan. Tagging and budgets show where money goes, and the first rightsizing review removes idle resources that had been running for months.

Before you start

  • Access to current cloud accounts and bills
  • A list of applications and owners
  • Compliance or data residency requirements
  • Target regions for production
  • Who will own the cloud after handover

What success looks like

Success is a cloud setup that is understood, reproducible, and affordable: every resource defined in code, environments clearly separated, access controlled through single sign-on, and a monthly bill that can be explained line by line. When something fails, it can be rebuilt from code rather than from memory.

A shared vocabulary for trade-offs

AWS's Well-Architected Framework organizes cloud design around six pillars: operational excellence, security, reliability, performance efficiency, cost optimization, and sustainabilitySource [1]. The names apply on any cloud, and they make trade-offs explicit: a design can be more reliable and more expensive, and everyone should know which way it leans.

Platform choices also carry ongoing obligations. The Kubernetes project maintains only its three most recent minor releases, each with about a year of patchesSource [2], so a cluster means a yearly upgrade. We recommend simpler managed platforms unless your scale and team make that work worthwhile.

Frequently asked questions

Is bare metal cheaper than cloud?
For steady, predictable workloads it can be. Cloud wins on elasticity and managed services.
Can you migrate us between clouds?
Yes, with a plan for data transfer, DNS cutover, and parallel running.
Do you support hybrid setups?
Yes, connecting on-premise systems to cloud services securely.
How do you control costs?
Tagging, budgets, alerts, rightsizing, and reserved capacity where usage is stable.
Who has access to our accounts?
Scoped roles for the people working on your project, removable at any time.
Which cloud should we choose?
The one that fits your team, region, and services. If you already use one, staying there is usually simplest. We'll compare options honestly.
Can you reduce our cloud bill?
Often, through rightsizing, removing idle resources, storage tiering, and commitments for steady workloads. We start with a cost review.
Do you support hybrid or on-premise setups?
Yes, when data residency, latency, or existing hardware make it sensible.
How do you avoid vendor lock-in?
By using portable building blocks such as containers, PostgreSQL, and Terraform where practical, and being deliberate about provider-specific services.
Can you work with our existing cloud account?
Yes. We start with a review and improve it in place, or plan a clean landing zone and migrate.
Can you help us move off a single large server?
Yes. We plan a phased move to managed services or containers, with a rollback at each step.
Do you help with cloud credits or contracts?
We can advise on architecture choices that affect commitments, but contracts stay between you and the provider.
Can you document our existing setup?
Yes. Importing existing resources into infrastructure as code is often the first step.
Do you work with Odoo.sh or Odoo Online?
Yes. We integrate them with your cloud for backups, monitoring, and connected services, and advise when moving to self-hosting makes sense.
Can you set up multi-region deployments?
Yes, when availability or latency requirements justify the extra cost and complexity.
What does a cloud review include?
An inventory of resources, a cost breakdown, security and access findings, and a prioritized list of improvements with rough effort for each.
Can you support a move from one cloud to another?
Yes. We plan the move service by service, prefer portable components, and run both environments in parallel until the switch is verified.
Do you document architecture decisions?
Yes. Key decisions are recorded with their reasons, so future changes can be made with the original trade-offs in mind.
Which cloud provider do you recommend?
The one that fits your team, region, and services. The design principles, such as the Well-Architected pillars, apply on any of themSource [1].
Do we need Kubernetes in the cloud?
Often not. Managed container platforms cover many needs without a cluster's upgrade cycleSource [2].

Sources

  1. [1]The pillars of the framework (AWS Well-Architected) (opens in a new tab)Amazon Web Services documentation
  2. [2]Releases (opens in a new tab)The Kubernetes Authors