How to Run a Ping Test: Server Latency and Packet Loss

A ping test is the quickest way to see how long it takes to reach a server and whether packets get lost along the way. But "I got 30 ms" on its own says very little: how many packets were sent, what was the worst time, was anything lost, and is the problem on your connection or somewhere on the path to the server? This guide walks through running a ping test properly on Windows, Linux and macOS, reading the result, using tracert and MTR to find where a problem starts, and putting together a report your provider's support team can actually use.

What Is a Ping Test and What Does It Measure?

Ping sends an ICMP "Echo Request" packet to a host and waits for an "Echo Reply". The time between sending and receiving is the round-trip time (RTT), shown in milliseconds. A ping test tells you four things:

• Reachability: does the host answer at all?
• Latency: how many milliseconds a round trip takes on average.
• Packet loss: what share of packets never got a reply.
• Stability (jitter): are the times consistent or do they jump around?

Ping does not measure bandwidth, page load time or application response time. If ping is low but your site is still slow, the bottleneck is most likely server resources or the application rather than the network. Our guide on finding server performance problems in ten minutes covers that side.

How to Run a Ping Test on Windows

Open Command Prompt by typing "cmd" in the Start menu. The default ping command sends only four packets, which is enough to confirm the server is up but not enough to judge quality. Send more packets for a meaningful sample:

ping -n 100 203.0.113.10
ping -t 203.0.113.10          (runs until you press Ctrl+C)
ping -6 -n 50 your-server.example   (test over IPv6)

When the command finishes, a summary at the bottom shows packets sent, received and lost, plus the minimum, maximum and average time. That summary is what you should focus on.

Because ping relies on ICMP, a server that blocks ICMP returns "Request timed out". In that case, testing the service port over TCP is more reliable. In PowerShell:

Test-NetConnection 203.0.113.10 -Port 3389
Test-NetConnection 203.0.113.10 -Port 443 -InformationLevel Detailed

The line TcpTestSucceeded : True means the port is open and the service answered. If ping fails but the TCP test succeeds, the server is running; only ICMP is filtered.

Running a Ping Test on Linux and macOS

On Linux and macOS, ping keeps running until you stop it, so it is practical to set the packet count up front:

ping -c 100 203.0.113.10
ping -c 200 -i 0.2 203.0.113.10     (200 packets, 0.2 s apart)
ping -6 -c 50 your-server.example   (on some systems: ping6)

The Linux summary looks like this:

100 packets transmitted, 100 received, 0% packet loss, time 99141ms
rtt min/avg/max/mdev = 28.412/29.087/34.906/0.913 ms

The mdev value shows how far individual times deviate from the average and is a handy jitter indicator. To check a service port, use nc:

nc -vz 203.0.113.10 22
nc -vz 203.0.113.10 443

How to Read a Ping Result: Average, Maximum and Jitter

Do not judge by a single number. Look at how the three values fit together.

What is a normal ping? Light travels more slowly in fiber than in a vacuum, adding roughly 1 ms of round-trip time for every 100 km of distance, before router hops and non-straight cable routes are counted. A few milliseconds within one city, tens of milliseconds within a country and well over a hundred between continents are all normal. We cover the effect of distance in more depth in our guide to server location and ping in Turkey.

Seeing the Path with Tracert and Traceroute

Ping shows you the outcome; tracert shows you the route. Every router a packet passes through on its way to your server is listed as a hop, which lets you see where latency jumps.

Windows:  tracert -d 203.0.113.10
Linux:    traceroute -n 203.0.113.10
Linux (over TCP 443, for paths that filter ICMP):
          sudo traceroute -T -p 443 -n 203.0.113.10

The -d and -n flags turn off name lookups so the test finishes much faster. Seeing * * * on some hops is normal: that router simply does not answer these probes, and traffic is not stopping there. If later hops respond, everything is fine.

On Windows, pathping combines tracert and ping: it maps the route first, then sends packets to each hop for a while and calculates loss. A full run takes a few minutes, but it is useful on a Windows machine where you cannot install MTR:

pathping -n 203.0.113.10

Testing Packet Loss with MTR: The Most Reliable Method

MTR runs traceroute and ping together, continuously. For each hop it shows packets sent, loss percentage and best, average and worst times in a single table. It has become the standard format for network reports sent to providers.

Ubuntu/Debian:     sudo apt install mtr-tiny
AlmaLinux/Rocky:   sudo dnf install mtr
macOS (Homebrew):  brew install mtr

Report mode (100 packets, with IPs and AS numbers):
sudo mtr -rwzbc 100 203.0.113.10

Over TCP 443:
sudo mtr -rwzbc 100 -T -P 443 203.0.113.10

The most common mistake when reading MTR

If one hop in the middle shows 20%, 40% or even 60% loss but the destination shows 0%, that is not real loss. Internet routers prioritize forwarding traffic and give limited resources to ICMP aimed at themselves, so loss at a single intermediate hop usually just means that router could not be bothered to reply.

Real problems have two signatures:

1. Loss starts at one hop and continues at a similar rate on every later hop and at the destination.
2. Latency jumps at one hop and stays at that higher level on every later hop.

A time that looks high at one hop but drops again further along comes from the same ICMP deprioritization and can be ignored.

When a Server Will Not Answer Ping: ICMP and Firewall Settings

Failing to ping a freshly installed Windows Server is very common, because the Windows firewall does not answer inbound ICMP echo requests by default. To let your monitoring or your own tests work, add a rule that allows only echo requests. In an elevated PowerShell window:

New-NetFirewallRule -DisplayName "Allow ICMPv4 Echo" -Protocol ICMPv4 -IcmpType 8 -Direction Inbound -Action Allow
New-NetFirewallRule -DisplayName "Allow ICMPv6 Echo" -Protocol ICMPv6 -IcmpType 128 -Direction Inbound -Action Allow

On Linux, ufw allows echo requests in its default rules. If you wrote your own nftables or iptables rule set, rate-limiting ICMP is healthier than blocking it outright:

sudo iptables -A INPUT -p icmp --icmp-type echo-request -m limit --limit 5/second -j ACCEPT
sudo iptables -A INPUT -p icmp --icmp-type echo-request -j DROP

Never block ICMP entirely on IPv6. IPv6 depends on ICMPv6 for neighbor discovery and path MTU discovery, and blocking it completely breaks connectivity in confusing ways. Limiting echo requests is enough.

If a firewall mistake locks you out, you can log in through the console in the Client Area on a VDS, or through the browser-based HTML5 KVM console on a dedicated server, and fix the rule.

Step-by-Step Troubleshooting for High Ping

1. Use a cable. Tests over Wi-Fi are the most common source of misleading results.
2. Check your local network. Send 100 pings to your router, usually the first hop. If you see loss or spikes even there, the problem is at your home or office.
3. Use a reference target. Ping a public DNS resolver such as 1.1.1.1 or 8.8.8.8 at the same time. If both targets suffer, the issue is your internet connection.
4. Keep the line quiet. A large download, backup or stream during the test will raise latency on its own.
5. Measure at different times. Compare morning and evening results; latency that climbs in the evening usually points to provider congestion.
6. Test the reverse direction. Internet traffic can take different routes out and back. Log in to your server and run MTR toward your own public IP; the problem may exist in one direction only.
7. Check the server itself. A saturated CPU or network interface delays replies. On Linux start with top and vnstat; on Windows, the Performance tab in Task Manager.
8. Consider an attack. Sudden, heavy loss can mean the server is under a traffic flood. Our guide to DDoS protection covers the warning signs.

MTU problems: ping works but some sites will not load

If small pings go through but larger packets stall, there may be an MTU mismatch on the path. Disable fragmentation and test packet sizes. 1472 bytes of payload plus a 28-byte header equals the standard 1500-byte Ethernet MTU:

Windows:  ping -f -l 1472 203.0.113.10
Linux:    ping -M do -s 1472 -c 5 203.0.113.10

If you get "Packet needs to be fragmented" or "message too long", reduce the size until you find the largest value that passes. This is especially revealing on VPN and tunnel connections.

Using Ping Tests to Choose a Server Location

Ping tests are useful before you buy, not only when something breaks. If most of your users are in Turkey, a server located there usually gives the lowest latency; for users across Europe, Germany or France is often the better fit. Keep these points in mind:

• Test from where your users actually are and from their providers, not only from your own desk. Different mobile carriers and home broadband providers can produce different numbers.
• Weigh stability as heavily as the average. For gaming, voice and remote desktop, a steady 25 ms beats a 15 ms connection that keeps spiking to 150 ms.
• For websites, latency multiplies because the TLS handshake and every new connection repeat the round trip several times.
• If some of your users are on IPv6, repeat the test with -6. VDS, dedicated and GPU servers in our Turkey location include a free /64 IPv6 block.

For a broader look at location choice, read our guide to choosing a data center location.

Preparing a Ping and MTR Report for Support

A ticket that just says "the internet is slow" tends to bounce back and forth several times. Sharing the following in your first message gets the problem located much faster:

• Your public IP address and internet provider.
• The date, time window and whether the problem is constant or intermittent.
• A ping summary of at least 100 packets toward the server (min/avg/max and loss).
• An MTR report toward the server (mtr -rwzbc 100), or pathping -n output on Windows.
• The same test toward a public reference target such as 1.1.1.1.
• If possible, a reverse MTR from the server back to your IP address.

Paste results as text rather than screenshots so values can be copied and compared. Support requests are accepted 24/7.

Related Guides

• Why Server Location in Turkey Matters: Ping and Performance
• Server Slowing Down? Find Performance Problems in 10 Minutes
• How to Choose a Data Center Location
• How to Protect Against DDoS Attacks
• VDS plans
• Dedicated servers in Turkey
• Virtual servers in Germany

Frequently Asked Questions