Viral benchmark: VPS A crushes B. In prod, API lags — high steal time, throttled CPU. Marketing: 4 vCPU / 8 GB guaranteed; terms: fair use and shared vCPU.
Guaranteed resources is cloud's most polysemous phrase.
The benchmark measured a sixty-second burst, not your nightly plateau load.
Decode the vocabulary
vCPU = hypervisor thread often. Dedicated CPU = premium. Burst credits then cap. Fair use = throttle on 24/7 load.
| Label | Meaning | Check |
|---|---|---|
| 4 vCPU | Thread quota | Sustained % |
| Guaranteed | VM allocation | Steal time |
| Dedicated core | Less overselling | 2–3× price |
Shared vs VPS vs dedicated
Shared = soft CPU. Cloud VPS = VM quota, variable neighbors. Dedicated = physical. Hetzner and OVHcloud document differently — same-load compare.
Test 24 hours
stress-ng, app load, steal time monitoring. Compare tool identical scenario.
The peak: legal floor not absolute performance
Batch workloads and CPU AI
Nightly training or batch jobs consume sustained CPU — exactly what fair-use clauses target. Before promising client SLA on a guaranteed VPS API, measure CPU over twenty-four hours at real load, not a one-minute test.
Compare Hetzner and OVHcloud with same script, window, instance size. Document steal time and warning emails. That is an honest short-list basis for non-brochure workloads.
Go further with the inquiry
Apply the article to your concrete case: what volume, what service level, what contractual constraints from clients or regulators? Write answers before opening the pricing page. Request written documents — full grid, DPA, certification scope, renewal terms — not homepage screenshots.
Compare at least two providers via the directory and compare tool on identical thirty-six-month assumptions. Share the short list with whoever signs the contract and whoever operates infrastructure daily. Archive terms dated at subscription so due diligence is defensible later.
Go further with the inquiry
Apply the article to your concrete case: what volume, what service level, what contractual constraints from clients or regulators? Write answers before opening the pricing page. Request written documents — full grid, DPA, certification scope, renewal terms — not homepage screenshots.
Compare at least two providers via the directory and compare tool on identical thirty-six-month assumptions. Share the short list with whoever signs the contract and whoever operates infrastructure daily. Archive terms dated at subscription so due diligence is defensible later.
Batch workloads and CPU AI
Nightly training or batch jobs consume sustained CPU — exactly what fair-use clauses target. Before promising client SLA on a guaranteed VPS API, measure CPU over twenty-four hours at real load, not a one-minute test.
Compare Hetzner and OVHcloud with same script, window, instance size. Document steal time and warning emails. That is an honest short-list basis for non-brochure workloads.
Batch workloads and CPU AI
Nightly training or batch jobs consume sustained CPU — exactly what fair-use clauses target. Before promising client SLA on a guaranteed VPS API, measure CPU over twenty-four hours at real load, not a one-minute test.
Compare Hetzner and OVHcloud with same script, window, instance size. Document steal time and warning emails. That is an honest short-list basis for non-brochure workloads.
Decide and move forward without blind spots
Start by framing your real scenario on one page: volumes, client constraints, required support level, and a thirty-six-month budget including domain, backup, and essential add-ons. Compare at least two hosts via the directory and compare tool with identical assumptions, archive pricing grids dated on subscription day, then have both the contract signer and daily infrastructure operator validate the short list.
Frequently asked questions
Guaranteed vCPU = physical core?
Rarely entry/mid tier. Often shared threads with limited burst.
Fair use on VPS?
Sustained load beyond threshold may throttle despite headline.
Guaranteed vs burstable RAM?
Allocated RAM on honest VPS; watch swap. Kubernetes: requests vs limits.
How to test?
stress-ng, real 24h load, steal time — same scenario at two hosts.
4 vCPU on the spec sheet? Run stress-ng two hours — before promising client SLA.