Network lag, high latency, and severe packet loss—when encountering these problems, many people's first reaction is, "Is there a problem with the server?" But the truth is often more complex: the problem might lie somewhere along the path from your home to the server, or it might be on the return path from the server to your home.

To accurately determine which link in the chain is causing the problem, you need to master a complete set of network testing tools. This article explains the usage of all commonly used network testing tools clearly, from the most basic Ping to the most professional Looking Glass—easy for beginners to understand, and even experienced users can use it to fill in any gaps in their knowledge.

I. Ping: The Most Basic but Also Most Easily Misunderstood Tool

Ping is the most basic and commonly used command in network diagnostics, and almost all operating systems have this tool built-in. It calculates round-trip latency and packet loss rate by sending ICMP probe packets to the target server and waiting for a reply.

Basic Usage:

In Windows systems, press Win+R to open the Run window, type cmd to enter the command prompt. In Linux or Mac systems, open the terminal.

# Windows (Continuous ping, press Ctrl+C to stop)

ping targetIP -t

# Windows (Stop after sending 4 packets)

ping -n 4 targetIP

# Linux/Mac (Stop after sending 100 packets)

ping -c 100 targetIP

How to interpret the results:

Ping results mainly focus on three metrics:

- Average latency: The round-trip time from you to the target server. Normal latency to Hong Kong servers is 30-60ms, and to the US West Coast is 130-180ms. A latency exceeding 300ms will significantly degrade the experience.

- Packet loss rate: How many of the sent packets did not receive a reply. Ideally, it should be 0%; anything above 2% should raise concerns.

- Latency fluctuations: If the latency fluctuates wildly, it indicates network jitter, which may affect real-time applications.

Limitations of Ping:

Ping can only tell you whether it's accessible and fast, but it cannot tell you where the slowness occurs. Furthermore, Ping uses the ICMP protocol, and many network devices prioritize or discard ICMP packets, potentially leading to discrepancies between Ping results and the actual TCP user experience.

II. Traceroute/Tracert: Breaking Down the Routing Path

If Ping only tells you "how far the destination is," then Traceroute tells you "which stations it passes through along the way."

Traceroute displays every routing node a data packet passes through from your device to the target server using a hop-by-hop probing method. Each hop displays the IP address and latency.

Basic Usage:

# Linux/Mac

traceroute target IP

# Windows (command name is tracert)

tracert target IP

How to Interpret the Results:

Each line represents a routing node, displaying its IP address and response time. If a line displays "***", it means that the node did not respond—this may be due to the device being configured not to respond to probes, not necessarily a network outage.

When to use it:

When Ping results show high latency or severe packet loss, Traceroute can pinpoint the problem at each hop. For example, if the latency suddenly spikes from 20ms to 200ms at a certain hop, the problem lies at that node.

III. MTR: A "combination" of Ping and Traceroute

MTR (My Traceroute) combines the advantages of Ping and Traceroute, displaying latency and packet loss rate for each hop simultaneously and continuously updating the data.

Linux/macOS Installation and Usage:

# Ubuntu/Debian

apt-get install mtr -y

# CentOS

yum install mtr -y

# Run MTR (persistent mode)

mtr target IP

# Run MTR (report mode, outputs results after sending 100 packets)

mtr -r -c 100 target IP

Windows users: WinMTR can be used, a graphical version of the MTR tool. After downloading, extract and run the program. Enter the target IP address in the Host field and click Start.

How to interpret MTR results:

MTR output contains multiple columns of data:

- Loss%: Packet loss rate for each hop

- Avg: Average latency

- Best/Worst: Minimum and maximum latency

Key judgment rule: If the packet loss rate is high for a particular hop, but the packet loss rate returns to normal for all subsequent hops, it indicates that the intermediate node has set ICMP rate limiting, not a real network problem. If the packet loss rate is high for a particular hop and all subsequent hops continue to experience high packet loss, then that is the real network bottleneck.

The core value of MTR: It can accurately pinpoint the "breakpoint"—whether it's congestion at the domestic operator's exit, a detour by the international backbone network, or a problem with the service provider's data center—it's all immediately apparent with a test.

IV. iperf3: Professional-grade bandwidth testing tool

Ping and MTR measure latency and packet loss, but low latency does not necessarily mean sufficient bandwidth. A node with 50ms latency but only 5Mbps bandwidth might perform worse for video playback than a node with 150ms latency but 100Mbps bandwidth.

iperf3 is currently the most authoritative network throughput testing tool, accurately measuring the actual bandwidth of TCP and UDP.

Installing iperf3:

# Ubuntu/Debian

apt-get install iperf3 -y

# CentOS

yum install iperf3 -y

Usage (requires two machines):

Start the service on the server:

iperf3 -s

Default listening port is 5201.

Perform tests on the client side:

# Test upstream bandwidth (client → server)

iperf3 -c serverIP

# Test downstream bandwidth (server → client), add the -R parameter

iperf3 -c serverIP -R

# Multi-threaded test (10 concurrent streams, more fully utilizing bandwidth)

iperf3 -c serverIP -P 10 -t 30

How to interpret the results:

iperf3 will display the transmission rate in real time and finally output the summary data. If the measured bandwidth is much lower than the nominal value you purchased, it indicates that the bandwidth may be falsely advertised or there may be issues with shared over-selling.

V. Looking Glass: "Looking Back" from the Server's Perspective

All the tools above measure "outward" from your device, but internet routing is asymmetrical—you go out via route A, but return via route B. You only test the outward journey; you have no idea if the return journey takes a detour.

Looking Glass (LG) solves this problem. It's an online network probing tool deployed by your service provider within their data center, allowing you to initiate tests from the server side to see the true return route.

What Looking Glass can test:

- Ping: Tests return latency from the data center to your IP.

- Traceroute/MTR: Shows which routing nodes the return route passes through from the data center to your IP.

- Speed ​​test file download: Verifies the server's actual downlink bandwidth.

How to use Looking Glass:

Most reputable VPS providers offer a Looking Glass entry point, usually located at the bottom of their website or in their help center. After opening the LG page:

1. Select the data center node you want to test.

2. Select the test type (Ping/Traceroute/MTR).

3. Enter your local IP address or any target IP.

4. Click Execute and wait for the results.

Verifying "Genuine/Fake Optimized Lines" with LG:

Taking a CN2 line as an example: If a large number of IP ranges starting with 59.43.xxx.xxx appear in the return route, it indicates that it is using a genuine CN2 GIA line. If the return path consists entirely of nodes starting with 202.97 (163 backbone network), then the so-called "optimized route" may only be one-way.

VI. Online Testing Platforms: No installation required, just open and use.

If you don't want to type commands in the command line, several online tools are very useful:

IT Dog (itdog.cn): Can simulate pinging a single IP from different regions and different operators (China Telecom, China Unicom, China Mobile) across the country simultaneously. Enter the target IP, select the node, and click test. Suitable for quickly determining whether "slow access in a certain region" is a widespread phenomenon.

Ping.pe: Performs Ping and MTR tests on the target IP from dozens of nodes worldwide. Green on the page represents smooth access, and red represents packet loss. Scrolling down reveals detailed MTR data.

FastSpeedTest: A completely free browser-based speed test tool, requiring no registration or installation, directly testing download speed, upload speed, latency, jitter, and packet loss rate.

Mastering this toolkit means that when you encounter network problems, you'll no longer be someone who just says "the network is good but laggy"—you'll be able to accurately pinpoint, "The problem is with the third hop; the latency jumped from 20ms to 200ms."

Jtti cloud servers all come standard with premium CN2 GIA lines and dedicated bandwidth, and have withstood testing and verification by all the tools mentioned above. Whether you want to verify network quality before purchasing using Looking Glass, or continuously monitor performance after deployment using MTR and iperf3, Jtti provides you with transparent, stable, and verifiable cloud server products.

Visit the Jtti website now and choose a cloud server that has withstood rigorous testing.