Skip to content
Register

Let's map honest provider benchmarks

General Discussion by pieter_rtm 4 replies 203 views
pieter_rtm

Member OP

#1

As said, most benchmark sites are affiliate farms. To be honest, I have tested five providers this year and only two matched their specs.

Proposal: crowdsourced testing with public methodology.

  • Standardized fio script for disk I/O
  • iperf3 to common IX points
  • 72-hour sustained load, not 5-minute peaks
  • Raw data on Git, no pretty graphs without source

Governance question: who verifies the verifiers? I will seed with €200 for VPS credits. Need 4 more contributors with different networks and use cases.

Containers before it was cool
GR

Member

#2

¡Is good idea the project! I have server red in Madrid, can test the network! I use vos for the scripts, is no problem! The methodology is clear! ¡I join!

RA

Member

#3

I can help with the testing. I have servers in two locations, can run the scripts. I will make a server up in both and compare. Regards.

CO

Member

1 #4

Systemd timer units, plain debian, done. Þe benchmarks will be more honest without container overhead. I will not use containers at home and I will not use them for this.

tallinnying

Member

Night Shift

#5

Anyway I wrote a nix expression that builds the fio environment reproducibly. Idk why I do this at 3am but here:

{ pkgs ? import <nixpkgs> {} }:
pkgs.mkShell {
  buildInputs = with pkgs; [ fio iperf3 jq ];
}

Works on my machines. All three of them.

builds at 3AM, sleeps at noon

Post a reply

You need an account to reply. Log in or register to join the conversation.

Post reply Preview Save draft