Azure foundations, the words I had to get straight first

One page, fourteen terms. This is the vocabulary everything else in Azure is built on, and most of it gets used loosely enough that two people can agree on a sentence while meaning different things. Simple words on purpose. The AWS equivalents are at the end, for anyone arriving from there.

Server

A computer with no screen and no keyboard, whose only job is to run applications. Your laptop is a server with extras.

• CPU, memory, storage. That is the whole list.
• Applications live on servers so that people other than the developer can use them.

Data centre

A room full of servers in racks, wired together with switches and routers, with power and cooling.

• Companies used to run their own. A team bought the servers, racked them and kept them alive.
• The public cloud is somebody else’s data centre, rented by the hour.

Private, public and hybrid cloud

Who owns the data centre and who is allowed to use it.

• Private: your organisation owns the hardware and only your organisation uses it.
• Public: Azure, AWS or Google owns it and anyone with an account uses it.
• Hybrid: some of your workload on your own hardware, some in the public cloud, connected together.
• Hybrid is not multi-cloud. Multi-cloud is using two public providers. Hybrid is owning some of it yourself.

Virtualisation

Splitting one physical server into several virtual ones that behave like separate machines.

• This is what makes cloud possible. Without it, one customer would need one physical server.
• A virtual machine in Azure is a slice of a real server in an Azure data centre. You never see the real one.

API

The way software talks to Azure without a person clicking.

• The portal, the command line and the templates all send requests to the same API.
• Everything you can click, you can automate, because the click was an API call anyway.

Availability zone

A data centre. That is the entire definition.

• In everyday speech it is a data centre. In cloud vocabulary it is an availability zone. Same building.
• Zones in one region are kept physically far apart, so a fire or a power failure in one does not reach the others.

Region

A named group of availability zones in one geographic area. East US, Australia East, Central India.

• Regions exist for two reasons. Latency: put the application near the customers, so a request from Perth does not travel to Virginia and back. Downtime: if a whole area loses power, another region is still up.
• Pick the region closest to your users. Everything else about resilience comes from using more than one zone inside it.
• Choosing a region does not make an application resilient on its own. Two copies in two zones with a load balancer in front does.

Load balancer

The thing in front of two copies of an application that decides which copy answers each request.

• If one copy is unhealthy, traffic goes to the other. Without it, a second copy is just a second bill.
• Users never reach a server directly. They reach the load balancer.

Scalability, elasticity, availability, disaster recovery

Four words that get swapped for each other in interviews.

• Scalability: the system can be made bigger. Up means a larger server; out means more servers.
• Elasticity: it gets bigger and smaller on its own, following demand. Scalability is the ability, elasticity is the automation.
• High availability: it keeps working when one part fails, because there is more than one part.
• Disaster recovery: it can be brought back after everything in one place fails, because a copy exists somewhere else.

IaaS, PaaS, SaaS

How much of the stack Azure runs for you.

• IaaS: Azure gives you a virtual machine. The operating system and everything on it are yours.
• PaaS: Azure runs the platform. You give it code or a database, and patching is not your job.
• SaaS: Azure, or someone, gives you a finished application. Outlook is SaaS.
• The useful question is not which letter, it is “what am I responsible for here?”

Subscription

The billing boundary. Everything you create belongs to exactly one subscription, and that is where the invoice goes.

• A free trial is a subscription. A company might have several, often one per environment.
• Resource groups live inside a subscription, never across two.

Service and resource

A service is what Azure offers. A resource is the thing you created with it.

• Virtual Machines is a service. Your VM called jenkins-01 is a resource.
• Storage Accounts is a service. pumawebdev is a resource.
• You are billed for resources, not for services.

Resource group

A mandatory container that every resource must be created into. Nothing exists outside one.

• One resource belongs to exactly one resource group. It cannot be in two.
• Delete the group and everything inside it is deleted. This is the feature, and it is also how a practice environment is cleaned up in one action.
• Group by lifecycle. Things that are created, managed and removed together belong together. payments-dev, payments-prod.
• Add an owner and an environment tag when you create it, so the purpose is still clear a month later.
• Access, cost and monitoring are all managed at the group. A new engineer gets access to the group once, not to twenty resources one at a time.
• The interview question is “how do you manage your Azure resources?” The answer starts with resource groups per project, or per project and environment if the company has one subscription.

Azure Resource Manager

The layer that receives every create, update and delete request, from any tool, and carries it out. Microsoft’s ARM overview is the page to read once.

• Portal, CLI, templates, Bicep and the SDKs all go through it. There is no other way in.
• It is why the order of creation is fixed. A resource group has to exist before a VM can be put in it, whichever tool you use.

If you know AWS

The resource group is the concept to sit with. Every other row is a rename. That one changes how you organise everything.

The order I check when something is not where I expect: which subscription am I in, which resource group did it land in, which region was it created in, and is it a resource I created or a service I am looking at.