Independent server testing

What will you run?

Benchmarks for VPS and cloud servers. Pick a workload and inspect the recorded measurements.

Compare VPS byin

Explore published provider reports

HetznerAkamai (Linode)VultrDigitalOcean
Same workloads. Reviewable evidence.Read the methodology
The benchmark explorer

Matched plans.
Real measurements.

6 reports from DigitalOcean, Vultr and Akamai (Linode).
Shared and dedicated CPU, matched plan sizes.

Published measurements
PostgreSQL in-memory dataset · higher is betterBar: median of recorded repetitions · highest repetition
ProviderTransactions / secAvg latency
DigitalOceans-2vcpu-4gbShared CPU · 2 vCPU · 3.8 GiBLondon, UK · 17 Oct 2026
1,479
2.70 ms
Vultrvc2-2c-4gbShared CPU · 2 vCPU · 3.8 GiBLondon, UK · 17 Oct 2026
2,073
1.93 ms
Akamai (Linode)g6-standard-2Shared CPU · 2 vCPU · 3.8 GiBLondon 2, UK · 18 Oct 2026
2,269
1.76 ms
DigitalOceanc2-2vcpu-4gbDedicated CPU · 2 vCPU · 3.8 GiBLondon, UK · 17 Oct 2026
1,612
2.48 ms
Vultrvoc-c-2c-4gb-50s-amdDedicated CPU · 2 vCPU · 3.8 GiBLondon, UK · 17 Oct 2026
4,171
0.96 ms
Akamai (Linode)g6-dedicated-2Dedicated CPU · 2 vCPU · 3.8 GiBLondon 2, UK · 18 Oct 2026
2,033
1.96 ms
2 reports selectedCompare selected
More
Latest published benchmark for each plan in the selected location.Individual observations, not provider rankings. No 24-hour consistency runs yet.
Behind the numbers

Every number.
Backed by a test.

Recorded hardware. Repeatable workloads.
Results you can inspect.

Read the methodology
Recorded configurationABCSeparate test recordsCPUMemoryDiskNetworkDatabaseStartRecorded test windowEnd
Published record
  • Report JSON
  • Test context
  • Private evidence
Sanitized report + context
01 /

Record the deployment

Keep provider, plan, region, observed hardware and agent version with the result.

02 /

Repeat the workload

Retain the recorded repetitions and their original units. Missing measurements do not become scores.

03 /

Publish the evidence

Connect each reviewed result to its test version, limitations and sanitized report.

A score is useful when you can explain it.Inspect a published test record

Methodology concept · process illustration, not measured chart data. Pilot results do not establish 24-hour consistency or physical-host independence.

Tutorials

Latest tutorials.

Practical guides to the decisions behind your infrastructure.

Browse tutorials
Web appsBackground jobsDatabasesAnalyticsShared CPUYour appDedicated CPUAlmaLinux · Latest tutorial

Run LiteLLM on AlmaLinux 10 with Docker

Install LiteLLM on AlmaLinux 10 with Docker, PostgreSQL, and a localhost-only web UI that you can open safely through an SSH tunnel.

Read the tutorial

More tutorials

Server Apps

Run LiteLLM on Ubuntu 24.04 with Docker

AlmaLinux

Set Up FlowiseAI on AlmaLinux 10 with Docker

Server Apps

Set Up FlowiseAI on Ubuntu 24.04 with Docker

Explore by topicLinuxDatabasesNetworkingPerformance
A few more details

Before
you choose.

A little context for a better decision.

Does shape.host sell hosting?

Shape.host publishes server benchmark evidence and technical tutorials to help readers inspect hosting infrastructure.

How do I compare different providers?

Inspect each deployment’s resources, architecture, region, workload settings and test date. Individual reports and their repeated measurements are not provider-wide rankings.

Can I inspect the test results?

Each published report includes the approved, sanitized JSON snapshot. Raw intake logs and private review evidence stay in administration.

Do pilot tests establish consistency?

A pilot measures its recorded test window. It does not establish 24-hour consistency or long-term reliability. Each report retains its review and limitations.

Find your next server.

Start with your workload. Compare the evidence.

Explore benchmarks