How to Test Server Network Speed: Methods, Metrics, and Best Practices

Key Network Metrics You Must Know

Before you start running tests, it’s best to ensure that you understand what you are measuring. Network speed is not a single metric. There are different variables, each of which is related to a different aspect of the network. Here’s a quick breakdown:

Let’s get into the first method of measuring these variables.

Method 1: Using Speedtest CLI

Speedtest CLI is a public internet performance monitoring tool available for various operating systems, including Linux distributions. It’s undoubtedly the most popular tool among DevOps because of its simplicity and precision combined.

Using Speedtest CLI, you can test download speed, upload speed, latency (ms), and any lost packets, providing you with a clear network picture.

Step 1: Install Speedtest CLI

To install Speedtest CLI, we’re going to use the official Ookla version.

Ubuntu and Debian

sudo apt-get install curl
curl -s https://packagecloud.io/install/repositories/ookla/speedtest-cli/script.deb.sh | sudo bash
sudo apt-get install speedtest

CentOS, AlmaLinux, Rocky Linux

curl -s https://packagecloud.io/install/repositories/ookla/speedtest-cli/script.rpm.sh | sudo bash

Alternatively, you can straightforwardly download the installer through the official speedtest.net website. The available Linux OS download files are i386, x86_64, armel, armhf, and aarch64.

Step 2: Basic Network Test

speedtest --accept-license --accept-gdpr

The tool will automatically select a nearby server and then measure all the basics, including latency, download speed, upload speed, and packet loss.

Here’s an example output:

• Latency: 1.95 ms
• Download: 871.14 Mbps
• Upload: 767.41 Mbps
• Packet Loss: 0.1%

This instantly shows you how fast and how consistent your Linux server network speed is. The latency, download, and upload tell us how fast the network is running, while the packet loss shows potential inconsistencies.

Here are a few more commands to try on Speedtest CLI:

To wrap Speedtest CLI up, we can say that it’s a surface-level tool, but it does offer everything you need for basic monitoring, logging, and parsing.

Method 2: Using iPref3 Testing

iPerf3 measures the raw throughput between two machines. It gives you precise control over streams, protocols, and direction. Use it to test internal traffic, cross-data-center links, private VLANs, or replication paths.

Unlike Speedtest, iPerf3 requires two servers. This means that you’ll need to install the tool on the two servers that you’ll be using to measure the connection.

Step 1: Install iPref3 

The installation of iPref3 remains simple:

Ubuntu or Debian

sudo apt update
sudo apt install iperf3 -y

CentOS, AlmaLinux, Rocky Linux

sudo yum install iperf3 -y

Step 2: Start the iPref3 Server

On the destination server, you need to start listening. It uses 5201. This means that if there is a firewall running, you must open this port.

First, start the listener:

iperf3 -s

Open 5201 (if needed):

sudo firewall-cmd --add-port=5201/tcp --permanent
sudo firewall-cmd --reload

Step 3: Run Test from the Client

Now we’re back on the client-server. This is where we’re going to start the text and observe the feedback from the destination listener.

From the source server:

iperf3 -c SERVER_IP

Note: Replace “SERVER_IP” with the IPv4 of the destination server!

The output should look something like this:

• [SUM] 0.00-10.00 sec 10.8 GBytes 9.29 Gbits/sec

Here we can see for how long the test has been running, and the exact network speed during this time. You can run the command multiple times, compare the speed, and analyze problems.

Here are some additional iPref3 commands you can use:

💡Quick Explanation:

• -P 4 runs four parallel connections. Increase the number on 10G, 25G, or 100G ports to better saturate the link.
• -R reverses the test direction to verify symmetric throughput.
• -u enables UDP mode, and -b 1G sets the target bandwidth. UDP testing reports jitter and packet loss.

What’s unique about iPref3 is that it provides depth. You have two servers, and you can analyze the network between them from A-Z. You can measure exactly how these servers communicate, and whether something is slowing them down.

See Also: Smart Routing Explained: Intelligent Network Optimization

Method 3: Using Natcat Testing

Unlike Speedtest CLI and iPref3, Netcat is a more raw way to measure TCP transfer between two Linux servers. At its core, it does not calculate bandwidth automatically. It measures how much data is transferred over a specific time frame.

What’s unique about Natcat is that it’s very light. DevOps teams keep it on their servers at all times as a quick way to test internal network connectivity.

Step 1: Install Natcat

As always, the installation can’t be any simpler:

Ubuntu or Debian

sudo apt install netcat -y

CentOS, AlmaLinux, Rocky Linux

sudo yum install nc -y

Step 2: Start Listener

Similar to iPref3, with Netcat, you must start the listener service on the receiving server:

nc -l -p 5001 > /dev/null

This opens port 5001 and discards incoming data. Don’t forget to open port 5001 if there is a running firewall on your receiving server:

sudo firewall-cmd --add-port=5001/tcp --permanent
sudo firewall-cmd --reload

Step 3: Send the Data

From your source server, send the data:

dd if=/dev/zero bs=1M count=1024 | nc SERVER_IP 5001

Here’s a breakdown of exactly what this does:

1. First, the “dd” will generate 1 GB of zero data.
2. bs=1M count=1024, which is exactly 1024 MB.
3. The data will stream through Netcat to the host.

In the meantime, the tool measures exactly what is going on. So, for instance, if this 1GB goes in for 1 second, you’ll get an output of about 8 Gbps.

You can also add time management for clearer results:

time dd if=/dev/zero bs=1M count=2048 | nc SERVER_IP 5001

In this example, 2GBs are transferred; if completed in 2 seconds, you get 15 gigabits = 8 Gbps.

See Also: How to Allocate an IP Address to a Virtual Server

Method 4: Using curl and wget

Unlike the other tools we’ve reviewed, using curl and wget, you can get accurate results of the real-world download performance. In this case, you’re not generating test data or measuring the speed between two servers; you’re connecting to actual HTTP or HTTPS endpoints for testing.

The method of using curl and wget is primarily used to verify CDN performance, create backup mirrors, or check the speed of a large file migration.

To perform the test, we’re going to need to start downloading a file.

Test Download Speed with wget

We’re going to download a 1GB test file from a URL of your choice:

wget -O /dev/null [INSERT DIRECT FILE DOWNLOAD URL HERE]

While downloading, you get live transfer speed in MB/,s and when the download is ready, wget will show you the total download speed, e.g, 196 MB/s.

If you want to convert it to Gbps, just:

• 202 MB/s × 8 = 1.61 Gbps

Quick Tip: You can use wget to test speeds at different times and identify congestion patterns.

Test Download Speed with curl

With curl, we’re pretty much doing the exact same thing with the URL:

curl -o /dev/null [INSERT DIRECT FILE DOWNLOAD URL HERE]

With curl, you need an extra command to display the average speed:

curl -o /dev/null -s -w "%{speed_download}\n" [INSERT DIRECT FILE DOWNLOAD URL HERE]

Note: To convert the output to megabytes, just divide by 125,000.

The best part about testing with wget and curl is that you can find URLs offering dummy files for testing purposes from different countries. This way, you can identify specific CDN issues that are location-based and may affect a good portion of your services. 

If you want to do further testing, you can try ServerMania Network Test Tool.

See Also: How to Configure Network Switches?

Need Consistent High-Speed Network Performance?

Testing is always the first step. If your current network performance is falling short, the right infrastructure can make an immediate difference.

Here at ServerMania, we offer robust network infrastructure through our public cloud (AraCloud) and dedicated servers. We know the true meaning of zero latency, no congestion, and 99.99% uptime SLA.

With ServerMania, you get:

1. Unmetered dedicated servers for predictable bandwidth usage
2. 10 Gbps to 100 Gbps unmetered servers for higher throughput
3. Data center locations across North America, Canada, and Europe

Curious to learn more about optimizing your network infrastructure? Don’t hesitate to book a free consultation with a network technician.

💬 If you want to get in touch with us immediately, contact our 24/7 customer support. We’re online and available to speak with you right now.