DigitalOcean’s appeal is that you can understand the whole product in an afternoon. The pricing is flat monthly, the plans are few, and there is no capacity planning ritual. The one decision that genuinely matters is shared versus dedicated CPU.
The plan families
| Plan | CPU | Suits |
|---|---|---|
| Basic (shared) | Shared vCPU | Development, low-traffic sites, staging, workers that idle |
| Premium Intel / AMD | Shared, newer silicon and NVMe | Same workloads, better per-core speed and faster disk |
| General Purpose | Dedicated | Production applications with steady load |
| CPU-Optimised | Dedicated | Build servers, encoding, CPU-bound APIs |
| Memory-Optimised | Dedicated | Caches, in-memory data, large databases |
| Storage-Optimised | Dedicated + large NVMe | Data-heavy workloads needing local disk |
Shared versus dedicated
Shared vCPU Droplets place you on a host with other tenants. For workloads that idle most of the time this is excellent value. Under sustained load you can experience contention, where your throughput varies for reasons that have nothing to do with your code.
The signal that you have outgrown shared CPU is variance: response times that fluctuate without a matching change in traffic. Move to a dedicated plan before you spend a week profiling your own application for a problem that is not in it.
Sizing method
- Measure memory working set first. Running out of RAM causes swapping, and swapping on a cloud disk is catastrophic for latency.
- Check CPU utilisation over a week, not an hour. Size for the busy period.
- Watch I/O wait. High I/O wait with idle CPU means you need faster storage, not more cores.
- Prefer scaling out across two smaller Droplets behind a load balancer over one large one, for availability as much as performance.
What the account tier controls
Plan choice and account limits are different things. The account caps how many Droplets can run concurrently. Our DigitalOcean listings cover four configurations including 3 and 10 Droplet tiers, and options with outbound SMTP on port 25 opened.
Size the tier against peak concurrent Droplets, including anything that exists briefly during a deploy. The full detail is in Droplet limits and port 25.
Managed services or do it yourself
Managed Databases cost more than the same engine on a Droplet you administer. They are usually worth it: backups, failover and patching are real work. Run your own only when you have a specific reason and someone who owns it.
Where DigitalOcean stops being the answer
There is no managed data warehouse and the machine learning tooling is thin. If analytics or ML is central, Google Cloud or AWS are the comparison. If you want the cheapest possible cores in Europe, see DigitalOcean vs Hetzner.
Common questions
Can I resize a Droplet?
Yes. CPU and RAM can be resized, and disk can be increased permanently but not shrunk.
Is bandwidth included?
Plans include a monthly transfer allowance, pooled across Droplets, with overage billed beyond it.
Do backups cost extra?
Automated backups are a paid add-on priced as a percentage of the Droplet cost. Snapshots are billed by stored size.