Cloud and VPS bills tend to grow quietly: an oversized server here, a forgotten snapshot there. Cloud cost optimisation starts with measuring what you actually use, then trimming waste without hurting performance.
Before you start
- Access to your provider's billing breakdown.
- A week or more of usage metrics from monitoring; see server monitoring basics.
Measure real usage
- CPU: average and peak. Mostly under 10–20% suggests an oversized server.
- RAM: look at used memory excluding cache (
free -h, the available column). - Disk: space used and I/O load (disk usage tools).
- Bandwidth: monthly transfer versus your allowance.
Right-size servers
- Move to a smaller plan when peaks leave lots of headroom; keep enough for traffic spikes and updates.
- Choose the right shape: memory-heavy apps need RAM, not more vCPUs.
- Consolidate several tiny servers onto one, using containers for isolation; see containers vs VMs.
Remove idle and forgotten resources
- Old test servers, unattached volumes and floating IPs.
- Snapshots and images nobody uses.
- Load balancers or databases left after a project ended.
Tag resources by project and owner so it is clear what can be removed.
Use better pricing models
- Reserved instances, savings plans or annual VPS billing for steady workloads.
- Spot/preemptible instances for interruptible batch jobs.
- Fixed-price VPS with included bandwidth instead of per-GB egress for traffic-heavy sites; see egress costs.
Optimise storage and traffic
- Move archives to cheaper object storage tiers with lifecycle rules.
- Set backup retention so old copies expire.
- Put a CDN in front of static content and compress responses.
Make software do more with less
- Caching (page cache, object cache, OPcache) often lets a smaller server handle the same traffic.
- Fix slow database queries before adding hardware.
- Autoscale only where traffic is truly spiky; see autoscaling explained.
Consider managed hosting for websites
For ordinary websites, shared or reseller hosting can cost less than running and maintaining your own VPS, with backups and security handled for you. See hosting types compared.
Frequently asked questions
How much headroom should I keep?
Enough for normal peaks plus a margin, often keeping average CPU well under 50% on single servers.
Will downsizing cause downtime?
Usually a reboot; disk downsizing is often not supported, so plan disk size carefully.
What is the most common waste?
Forgotten resources and oversized servers chosen "just in case".