Differences in Ookla test servers

TL;DR

When we tested local Ookla test servers (all located in Amsterdam, N=40) with the Ookla Speedtest, we conclude that 5% of the test servers deliver speeds that are too low, 73% as expected and 22% as too high.

The worst performing test server delivers an internet speed that is more than 5% slower than the median.

The Ookla test server used can have a significant impact on reported speed.

 

Contents

  1. Introduction
  2. Method of measurement
  3. Test servers to test
  4. The measurements
  5. Conclusions
 

Introduction

Earlier we concluded that not all Ookla test servers are the same. We started that test because the study A Comparative Analysis of Ookla Speedtest and Measurement Labs Network Diagnostic Test (NDT7) (PDF) states, among other things, the following: (...) For Ookla, the choice of test server can significantly affect the reported speed. Tests using certain Ookla servers systematically report speeds 10% lower than other servers (...).

Ookla claims to have over 15,000 test servers. With this sample we try to find out if we can find out how many test servers deliver "the correct speed" (which we still think should be 100.8 Mbps).

 

Method of measurement

For this test we use the original Ookla Speedtest, and we will run one speed test per test server (hosted in Amsterdam).

One test per test server, because our previous test gives the idea that the standard deviation is so small that it is not necessary to run multiple tests per test server.

Amsterdam because our previous test showed that 3 out of 10 local test servers were located in Amsterdam, and there are a lot of test servers in Amsterdam.

To get a good insight, we group the test servers at (rounded) 0.1 Mbps.

 

Test servers to test

We test all servers located in Amsterdam.

 

The measurements

  1. Rounded 95.2 Mbps
    1. CLOUD LEASE: 95.17 Mbps
  2. Rounded 98.8 Mbps
    1. Kamatera, Inc: 98.75 Mbps
  3. Rounded 100.4 Mbps
    1. Odido: 100.37 Mbps
  4. Rounded 100.5 Mbps
    1. bouwhuis.network: 100.45 Mbps
    2. drie-host: 100.45 Mbps
    3. ServerRoom.net: 100.45 Mbps
    4. TNGNET B.V.: 100.45 Mbps
    5. Eranium B.V.: 100.47 Mbps
    6. NFOrce Entertainment B.V.: 100.48 Mbps
    7. Sharktech Inc.: 100.48 Mbps
    8. Usenet.Farm: 100.48 Mbps
    9. Clouvider Ltd: 100.49 Mbps
    10. HyperFilter DDoS Protection Solutions: 100.49 Mbps
    11. Melbicom: 100.49 Mbps
    12. WD6.net: 100.50 Mbps
    13. dstny: 100.51 Mbps
    14. fdcservers.net: 100.51 Mbps
    15. Solcon Internetdiensten B.V.: 100.51 Mbps
    16. BlackHOST Ltd.: 100.52 Mbps
    17. CSL Networks: 100.52 Mbps
    18. Hivelocity: 100.52 Mbps
    19. Labixe Ltd: 100.52 Mbps
    20. MIRHosting: 100.52 Mbps
    21. Vancis: 100.53 Mbps
    22. Qonnected B.V.: 100.53 Mbps
    23. I4 Networks: 100.54 Mbps
  5. Rounded 100.6 Mbps
    1. NewsXS B.V.: 100.57 Mbps
    2. XS News B.V.: 100.59 Mbps
  6. Rounded 100.7 Mbps
    1. WARIAN: 100.65 Mbps
    2. RETN: 101.71 Mbps
    3. Voiped Telecom: 101.71 Mbps
  7. Rounded 101.4 Mbps
    1. Global Layer: 101.41 Mbps
  8. Rounded 101.9 Mbps
    1. VDSina: 101.85 Mbps
    2. Fiber NL B.V.: 101.89 Mbps
    3. PhoenixNAP Global IT Services: 101.89 Mbps
    4. Asimo Networks B.V.: 101.90 Mbps
    5. 31173 Services AB: 101.91 Mbps
    6. Hack The Box Ltd: 101.91 Mbps
    7. Nextpertise: 101.91 Mbps
    8. AVUR AS: 101.92 Mbps
 

Conclusions

When we considered everything below the 100 Mbps as too low and every thing above the 101 Mbps as too high, then there are:

  1. 2 test servers which deliver speeds that are too low (5%)
  2. 29 test servers which delivers speeds as expected (73%)
  3. 9 test servers which delivers speeds that are too high (22%)

Most test servers deliver 100.5 Mbps.

The lowest speed measured is 95.2 Mbps. This 5.3% is lower than the median of 100.5 Mbps.