Decision model
Use this order: workload type -> traffic profile -> operational ownership -> compliance needs.
Quick guidance
- Shared: brochure sites, small CMS, low ops overhead
- VPS: custom stacks, root access required, moderate scaling
- Dedicated: high isolation/performance workloads
- Managed: you want operations, patching, scaling handled
Capacity planning baseline
- Start conservative and scale after real traffic metrics.
- Capture p95 latency, CPU saturation, memory pressure, and disk I/O trends before upgrading.
Upgrade trigger examples
- sustained CPU > 70%
- memory pressure with swap growth
- response-time regressions during peak windows