Skip to content
~/leonardo.casoni

cloud operations

What does a Cloud Operations team do?

A quick guide to the basic concepts and what is usually expected from the people who run a company’s cloud.

In one sentence

A Cloud Operations team keeps the services running in the cloud (AWS, Azure and the like): reachable, fast, secure and not costing more than necessary. If developers build the products, Cloud Operations makes sure there is a solid home for them to live in, day and night.

The line between Cloud Operations, DevOps and SRE is blurry and changes from company to company. More than labels, what matters is the idea: the people who build and the people who run work together, automate as much as possible and learn from incidents.

What is usually expected

Responsibilities change from team to team, but these are the ones I find almost everywhere.

Availability and reliability

Monitor systems, define what “working” means (SLAs and SLOs) and act before users notice.

Incident management

Respond to production problems in an orderly way, even at night, and learn from each incident.

Security

Manage access and permissions, updates, vulnerabilities and secure configurations, together with the security team.

Costs (FinOps)

Understand where the cloud money goes, switch off what is not needed and size resources properly.

Automation

Describe infrastructure as code (for example with Terraform) and automate repetitive tasks, to reduce errors and time.

Changes and releases

Get changes into production safely, with CI/CD pipelines and a clear change process.

Backup and disaster recovery

Know how to restore services after a serious failure, and check that it really works.

Migrations and modernization

Move and renew applications and data between environments and providers, with as little risk as possible.

Performance and capacity

Forecast growth, scale at the right time and find bottlenecks.

Support and documentation

Help developers use the platform well and leave runbooks and guides anyone can follow.

How the work looks in practice

  • On-call rotations: someone is always ready to step in. With teams across time zones you can use a Follow the Sun model, where the shift is handed over following the sun, so nobody stays on duty at night for long periods.
  • Runbooks: step-by-step instructions for the most common problems, written to be followed even under stress.
  • Useful alerts: few, clear, and requiring real action. Too many alarms only create noise.
  • Blameless reviews: after an incident you look for what in the system allowed the mistake, not who made it.
  • Regular reports on costs, security and reliability, to decide with data and not by gut feeling.

The words you will hear most

SLA
Service Level Agreement: the commitment to the customer on a level of service, for example 99.9% availability.
SLO
Service Level Objective: the internal target, usually more ambitious than the SLA.
SLI
Service Level Indicator: the real measurement, for example the percentage of successful requests.
MTTD / MTTR
Mean time to detect a problem and mean time to restore the service.
RTO / RPO
How long you can stay down and how much data you can afford to lose after a disaster.
IaC
Infrastructure as Code: infrastructure described in code files, versioned and repeatable.
Observability
The ability to understand what happens inside a system through metrics, logs and traces.
FinOps
The practice of managing and optimizing cloud costs together with engineering and business.
Runbook
A written procedure to handle or fix a known situation.
Post-mortem
The review done after an incident to understand what happened and what to improve.
Shared responsibility
The cloud provider secures the cloud; you secure what you put in it: configurations, access and data.

If you want to start

  • Linux and shell: the basis of almost everything.
  • Networking: DNS, IP addresses, firewalls and load balancers.
  • A cloud provider: pick one (AWS, Azure or Google Cloud) and start with the foundational certifications.
  • Infrastructure as Code: Terraform is a good starting point.
  • Monitoring and logs: learn to read metrics and error messages.
  • A project of your own: build it, break it, restore it and document what you learned.

This is an introductory guide, written the way I understood things while working in the field. Every company organizes itself differently: take it as a starting point, not as a manual.

Keep reading