How to choose the location of a virtual machine

When you create a virtual machine (VM), you choose where it runs. Most people pick the location closest to them. That is a good first guess, but the closest location on the map is not always the fastest. This page explains what decides the speed, and how to measure it yourself.

Near on the map is not near on the network

Your data does not travel in a straight line. It passes through a chain of networks owned by different companies, and the route is often not the direct one.

Two servers at the same distance from you can differ a lot: the route to one is direct, the route to the other goes through another city or country first. What counts is the quality of the path between you and the server, and you cannot see that on a map.

The three things that decide speed

	What it is	What you notice
Latency	The delay before an answer arrives, in milliseconds	High latency makes everything feel slow to react, even on a fast connection
Bandwidth	How much data can move per second	Matters for large downloads and video, less for small requests
Packet loss	Data that goes missing on the way and is sent again	Even 1% or 2% makes a connection feel unreliable

The route affects latency and packet loss the most.

Your internet provider changes the answer

The same location can be fast for one person and slow for another in the same city, because they use different internet providers. Each provider has its own routes and its own agreements with other networks.

So another person’s advice on the best location is of limited use. Measure from your own connection, and where you can, from a connection like the one your users have.

Measure before you decide

To measure a location you need the address of a machine there. The sure way is to create the smallest VM in each location you are considering, test its IP address, and delete the ones you do not keep. A VM is charged by the hour, not by the month. Delete the test VMs afterwards: a stopped VM is still charged.

Latency and loss: ping

ping -c 10 YOUR_VM_IP

-c 10 sends ten packets and stops. On Windows, the command is ping -n 10 YOUR_VM_IP.

Read the last two lines:

• Packet loss should be 0%.
• The average time (avg) is the latency. Lower is better, and it should not jump around.

The route: mtr

mtr shows every step on the way to the server, and where delay or loss begins.

mtr -rwc 50 YOUR_VM_IP

-r prints a report and exits, -w shows full host names, and -c 50 sends 50 rounds of packets.

If mtr is not installed: apt install -y mtr-tiny on Ubuntu and Debian, dnf install -y mtr on Fedora, AlmaLinux and Rocky Linux, brew install mtr on macOS (then run it with sudo). On Windows, tracert YOUR_VM_IP shows the route without the loss figures.

Each line of the report is one step. Loss at a single step in the middle that does not continue to the last line is harmless: that router answers test packets slowly. Loss or delay that begins at one step and stays until the last line is a real problem on the path.

How to choose

1. Start with the locations closest to the people who will use the VM.
2. Pick two or three.
3. Measure each with ping and mtr, more than once and at different times of day.
4. Choose the one that is best every time, not the one that is nearest.

If your users are spread over several regions, choose the location that works best for the largest group.

Locations on Doprax

The locations differ per provider. Providers and locations lists them all. On the Create Virtual Machine page you choose the provider first and then one of its locations.