Testing your VPS network speed

What this is#

How to measure your VPS's actual throughput, and how to read the numbers before concluding something is slow. (If your problem is lag, jitter, or packet loss rather than transfer speed, that's the network problems guide, the two pair well.)

What to expect before you test#

• Your VPS sits on a fast 10 Gbps uplink, but a speed test never measures "the port" in isolation, it measures the whole path to whatever server you test against, and the slowest link on that path wins.
• A single TCP connection slows down with distance. That's TCP physics (latency limits how fast one stream can ramp), so one stream to a nearby server can pull gigabits while one stream across the Atlantic pulls far less, on a perfectly healthy connection. Multiple parallel streams show the real capacity.
• Testing from home measures your home line. If your home plan is 100 Mbit, every download from your VPS tops out there, that's not the VPS.

Method 1: Speedtest CLI on the VPS#

Ookla's official CLI is the quickest meaningful test. Install it on the VPS (their install instructions cover every distro), then:

speedtest

It picks a nearby server and reports down/up/latency. Two tips: test against a nearby server first (that's the fair test of the VPS's connection), and if you want a specific target, list servers with speedtest -L and pick one with -s ID. Skip the older community speedtest-cli Python script, it underreports at high speeds.

Method 2: our Looking Glass test files#

The Looking Glass publishes 100MB, 1GB, and 10GB test files for each of our locations (for example https://lg.dallas.vpsdime.com/1GB.test, pick your location on the page).

• From your home browser: downloading one tests the path from our network to you, this is the number to check before ordering, or when "downloads from my VPS are slow at home".
• From a VPS: wget -O /dev/null https://lg.dallas.vpsdime.com/1GB.test tests between our locations or from another provider's box.

One caution: test downloads count against your VPS's monthly traffic allowance like any other transfer, prefer the 1GB file for routine checks and save the 10GB one for when you need it.

Method 3: iperf3 between machines you control#

For measuring one specific path properly (VPS to VPS, VPS to office), iperf3 is the gold standard, run both ends yourself:

iperf3 -s                          # on the receiving machine
iperf3 -c OTHER.HOST -P 4          # from the sending machine, 4 parallel streams
iperf3 -c OTHER.HOST -P 4 -R       # same, reversed direction

-P 4 (or 8) shows path capacity rather than single-stream TCP behavior, and -R matters because the two directions can differ. Open TCP port 5201 in your own firewall on the server side for the test, and close it after.

Reading the results#

• One stream slow, parallel streams fast: normal TCP-over-distance behavior, not a fault. Judge capacity by the parallel number.
• Slow to one destination, fast to others: a path/peering issue toward that destination, run mtr both ways and send us the traces.
• Slow everywhere, tested from home: test from the VPS itself (Method 1); if the VPS tests fine, the bottleneck is your local line.
• Genuinely slow from the VPS to nearby servers too: now it's worth a ticket, include the speedtest/iperf3 outputs, which server you tested against, when, and an mtr.

Related questions#

• "How do I run a speed test on my VPS?"
• "Why is a single download from my VPS slow?"
• "Do you have test files I can download?"
• "How do I test bandwidth between two servers with iperf3?"
• "Does speed testing use up my traffic allowance?"