Cloud computing
Practical writing on cloud providers, infrastructure, deployment and cost, from the team behind the BuyCloudExpress catalogue.
How the AWS Free Tier actually works
The AWS Free Tier is three different things wearing one name. Knowing which parts expire, which never do, and which quietly bill you is the difference…
EC2 instance families, decoded
EC2 instance names look like line noise until you learn the pattern. Once you can read them, choosing a shape becomes a two-minute decision instead of…
Azure tenants, subscriptions and resource groups: how the hierarchy fits together
Azure's account hierarchy has more layers than most clouds, and getting it wrong is expensive to undo. Here is what each layer is actually for.
Spot VMs on Google Cloud: when the discount is worth it
Spot instances trade certainty for a large discount. The question is never whether they are cheaper, it is whether your workload can survive being interrupted with…
Oracle Cloud Always Free: what you actually get
Oracle's permanently free tier is unusually generous, particularly its Arm compute allowance. It also has real constraints that decide whether it fits your project.
Oracle Autonomous Database, in plain language
Autonomous Database automates the work a database administrator would otherwise do. Knowing which parts it automates, and which it does not, tells you whether it replaces…
How to read an uptime SLA properly
A 99.99% SLA is a billing agreement, not a promise about your application. Reading one correctly means knowing what it measures, what it excludes, and what…
Choosing a data centre location that matches your users
Region choice sets a latency floor you cannot optimise past, and a legal context you cannot ignore. Both are decided at provisioning time and expensive to…
Data sovereignty in Europe: what it means for your architecture
Sovereignty is not the same as residency, and confusing the two produces architectures that satisfy neither. Here is the distinction and what it changes.
The OVHcloud product range, decoded
OVHcloud sells several overlapping compute products with similar names. Knowing which is which turns a confusing catalogue into a short decision.
Designing for a region outage without rebuilding everything
Full multi-region redundancy is expensive and hard. There is a great deal of useful resilience available well before that point, and most teams skip straight past…
When Kubernetes is the wrong answer
Kubernetes solves real problems very well. Adopting it before you have those problems trades a small amount of deployment friction for a large amount of permanent…