Requirements and sizing
Working out CPU, memory and storage from real usage data rather than a guess, and choosing between shared, VPS, dedicated and cloud on the evidence.
Servers, Hosting & DevOps
Cloud hosting & VPS setup
Hosting decisions are usually made once and then live with you for years, so they are worth a little thought. The two failure modes are equally common: paying for cloud infrastructure sized for traffic that never arrived, and running a growing business on shared hosting that cannot support it.
What I try to establish first is the actual shape of the workload, traffic pattern, storage growth, how much downtime genuinely costs, and then pick the smallest arrangement that satisfies it, with a clear path to scale up.
How I work on it
Process
Adjusted to the job, but this is the shape it usually takes.
-
Measure the current load
Existing traffic, peaks, database size and growth rate. Sizing without this is guesswork in both directions.
-
Choose for the next year, not five
Cloud infrastructure is easy to resize. Over-provisioning "for growth" is one of the most reliable ways to waste money.
-
Build repeatably
Provisioning captured in a script or Ansible playbook so the box can be rebuilt rather than resurrected.
-
Review after a month
Real utilisation data almost always suggests a size adjustment, usually downwards.
Questions
Things people ask
Not inherently, and often not for the money. AWS is worth it when you need its managed services, regional coverage or compliance posture. For a single application with predictable traffic, a straightforward VPS is frequently faster to run and several times cheaper.
A brochure site with modest traffic runs comfortably on 2 vCPU and 4 GB with proper caching. WooCommerce with a large catalogue and logged-in users bypassing the page cache needs meaningfully more, because caching helps far less once every request is personalised.
If nobody on your side wants to be responsible for updates and alerts, yes, genuinely. Managed hosting costs more per month and less in incidents. I would rather set you up somewhere sustainable than sell you a server you will not maintain.