The Essential Guide to Cloud & IT

Points clés

• Sort every new service into compute, storage, networking, identity, operations or cost, then ask which of the other five it will touch.
• Whatever the service model, your data and who can reach it stay your responsibility, so find the shared line before you act.
• Set a recovery point and a recovery time target for important data, then prove your backups by restoring one on a schedule.
• Give every person and program the smallest set of permissions for the shortest time, and require multi-factor sign-in for anything powerful.
• Tag resources with an owner, set a budget alert and review the biggest costs monthly, because forgotten resources cause most surprise bills.

À l'intérieur de ce bit

1. Six ideas that hold the field together

   Cloud and IT can look like a thousand product names, but nearly every exam question and workplace ticket sits on six ideas. Compute is the processing power that runs programs.

2. Service models and the shared responsibility line

   Every cloud arrangement splits the work between you and the provider, and the split depends on the service model. With infrastructure as a service, you rent virtual machines and handle the operating system, updates, software and data yourself.

3. Data: recovery targets before storage types

   Storage comes in three broad shapes. Block storage behaves like a disk attached to one server.

4. Networks and identity: the two front doors

   Most serious incidents enter through one of two doors: the network or an identity. On the network side, a virtual network is your private space inside the provider, divided into subnets.

5. Reliability: design as if parts will fail

   Providers organise their data centres into regions, which are geographic areas, and availability zones, which are separate buildings within a region with their own power and cooling. Running copies of a service in two zones turns a failed building into a blip rather than an outage.

6. Cost and priorities: what to learn and ignore

   Cloud billing is pay as you go: you pay for what runs, often by the second, hour or gigabyte. That is a gift for experiments and a trap for forgetful teams.