Optimize Network Latency in Vietnam: VNPT vs Viettel vs FPT vs CMC

The first thing to check when your VPS in Vietnam feels slow is not the CPU or the disk, it is the route between your users and the server. Vietnam has four major upstream providers: VNPT, Viettel, FPT Telecom and CMC Telecom, and the path your traffic takes through one of them can add 5 ms or 50 ms depending on where your users sit. This post compares their domestic routing, explains what "network latency in Vietnam" actually means on the ground, and gives you the exact commands to measure it before you choose a provider for your Linux VPS.

Why domestic latency in Vietnam varies by provider

Vietnam does not have one national backbone. Each telecom operator runs its own physical fiber network, and interconnection between them happens at a limited number of IXP (Internet Exchange Point) locations in Hanoi and Ho Chi Minh City. If your server is on a Viettel line and your user is on VNPT, the traffic has to cross the exchange, and that hop is where the extra milliseconds appear.

This is why two VPS servers in the same datacenter can have very different response times for the same user. One is on the VNPT network, the other on Viettel, and the user's ISP decides which path to take. The result is a routing table that changes by region: a user in Da Nang on FPT gets a fast path to a server on the FPT network, but a slow one to a server on CMC Telecom.

So the practical rule for anyone running a VPS for Vietnamese users is simple: match your upstream to your audience. If most of your users are on Viettel fiber, a server on the Viettel network will almost always respond faster than one on VNPT, regardless of the hardware specs.

VNPT vs Viettel vs FPT vs CMC: what each network is best at

Here is how the four major providers compare for domestic routing in 2026:

This is not a quality ranking. Viettel is not "better" than CMC, it is different, and the right choice depends entirely on where your users sit. A B2B SaaS serving corporate clients in Hanoi might see better results on CMC Telecom, while a consumer app popular with students on Viettel mobile needs to be on the Viettel network.

How to measure network latency to a VPS in Vietnam

Do not guess. Measure before you buy. The tools are free, and you should run them from the locations where your actual users sit, not from your own office overseas.

Step 1 - Run a basic ping test

Ping gives you a quick round-trip time (RTT) number. Run it from a machine close to your users, or ask a friend in Vietnam to run it for you:

ping -c 100 -i 0.2 YOUR_SERVER_IP

The -c 100 flag sends 100 packets and -i 0.2 sets a 200 ms interval. The output ends with a summary that shows min, avg and max latency. A healthy domestic ping on fiber is typically under 10 ms within the same city and under 20 ms between Hanoi and Ho Chi Minh City. If you see an average above 50 ms, the route is taking extra hops.

Step 2 - Trace the route with MTR

Ping tells you the total time, but not where the time goes. MTR combines ping and traceroute to show every hop along the path. On Ubuntu or Debian, install it with:

sudo apt update
sudo apt install mtr-tiny

Then run it against your server:

mtr -rw -c 100 YOUR_SERVER_IP

The output shows each hop with packet loss and latency per hop. Look for three things. First, a high latency jump at the hop where traffic crosses from your ISP to the server's network, that is the IXP handoff. Second, packet loss on any hop, even 1% loss matters for TCP throughput. Third, a route that goes through an unexpected city, that adds 10-15 ms for no reason.

Step 3 - Test at peak hours

Domestic bandwidth in Vietnam, often 100 Mbps or 1 Gbps on the port, is shared and congestion appears at night. Run the same MTR test at 9 PM local time on a weekday and compare it to the morning result. If the evening latency is double the morning latency, the provider has an oversubscribed exchange. This is a common failure mode with cheaper plans.

Here is the key takeaway: a single ping test is not enough. You need to measure from the user's location, at peak hours, and compare a few candidate servers on different networks before committing.

Choosing an ISP for your VPS based on your users

Once you have the measurement data, the decision becomes straightforward. Match the server's network to the majority of your user base:

• Consumer app with national reach: choose a provider with broad peering like Viettel or FPT Telecom, then verify with MTR from two or three cities.
• Enterprise SaaS for Vietnamese companies: consider CMC Telecom for its Tier 3 datacenters (Tier 3 datacenter Vietnam) and business-grade routing, and verify that your target companies are on a compatible network.
• Niche audience in one city: pick the dominant ISP in that city. In Da Nang, for example, a large share of residential fiber is on VNPT and Viettel, so test both.
• International users accessing from overseas: domestic routing matters less, focus on the provider with the best international transit, often FPT Telecom, and test from your overseas location.

One common mistake is assuming all VPS providers in Vietnam use the same upstream. They do not. A WordPress VPS from one provider might sit on a Viettel line, while a similar product from another provider uses CMC. Ask the provider directly which upstream they use, or check the AS number (AS7552 for Viettel, AS45899 for VNPT, AS18403 for FPT, AS45903 for CMC Telecom) by looking up the IP address.

Practical steps to reduce latency for your VPS users

Even with the right provider, you can squeeze out more performance. These are the adjustments that actually move the numbers:

Deploy in the right region

Vietnam has two main peering hubs: Hanoi in the north and Ho Chi Minh City in the south. If your users are concentrated in the south, a server in Ho Chi Minh City is faster, and vice versa. For national reach, test both and pick the one with the lower average across your user locations.

Use a CDN or a reverse proxy

A CDN with Vietnam PoPs (points of presence) caches static content closer to the user and saves the round trip to your origin. For dynamic content, an nginx reverse proxy can terminate TLS and serve cached responses, which reduces the perceived latency even if the network path stays the same.

Enable TCP BBR

The default TCP congestion control in Linux is fine, but BBR often performs better on high-bandwidth, high-latency paths. To enable it on a modern kernel:

sudo sysctl -w net.core.default_qdisc=fq
sudo sysctl -w net.ipv4.tcp_congestion_control=bbr
sudo sysctl -p

Verify that BBR is active:

sysctl net.ipv4.tcp_congestion_control

The output should show bbr. This is a low-risk change that helps mostly on longer paths, like from Vietnam to international destinations.

Keep DNS resolution fast

A slow DNS lookup adds 20-50 ms to every first request. Run your own resolver on the VPS with systemd-resolved or point your DNS records to a fast nameserver like Cloudflare. Test with dig:

dig +time=2 +tries=1 example.com

Look at the Query time in the output. Under 30 ms is good for a domestic resolver.

越南 VPS 应选择与用户相同 ISP 的网络，以减少本地路由延迟。

Choose a VPS on the same ISP network as your users to minimize domestic routing latency.

How to pick a VPS provider that lets you choose the network

The ideal setup is a provider that lets you pick the upstream, or at least tells you which one you are getting. With a Linux VPS, you want full root access (sudo) so you can run MTR, tune sysctl, and install tools without restrictions. Some providers force you through a NAT or a shared gateway, which breaks MTR accuracy at the first hop.

Before you commit, ask the provider three questions:

1. Which upstream network does the server use, VNPT, Viettel, FPT or CMC?
2. Is a dedicated IPv4 address included, or is it shared?
3. Can you access the server with full root and install custom kernel modules?

A provider that answers these clearly is more likely to understand domestic routing than one that gives a vague reply. You can also do a quick IP WHOIS lookup before paying to confirm the upstream AS number.

Test multiple providers before committing

Run the same MTR test on a candidate VPS from each major network at peak hours. Keep the results in a table:

(These numbers are illustrative, not real measurements. Yours will differ. The point is the comparison method.)

If you are evaluating thueVPS, they run on NVMe storage with a dedicated IPv4 in Vietnam and KVM virtualization, which means no NAT and full root access, so your MTR results reflect the real network path.

Troubleshooting: why is my VPS still slow after switching?

You changed providers, measured, and still see high latency. Here are the usual causes and the diagnostic commands to confirm them:

Cause 1 - The user's own ISP is the bottleneck

Your server is fast, but the user's home connection is oversubscribed. There is nothing you can do about their local loop. Diagnose by running MTR from a second location, if the second location shows low latency, the problem is on the user's side, not yours.

Cause 2 - TLS handshake overhead

Adding TLS adds two round trips before the first byte. Use TLS 1.3 and session resumption to cut this down. Test the handshake time with:

curl -w "TLS handshake: %{time_appconnect}s\n" -o /dev/null -s https://your-site.com

If the handshake takes more than 100 ms, consider enabling TLS 1.3 and ssl_session_cache in nginx.

Cause 3 - The IXP is congested at peak hours

If your MTR shows the latency jump only in the evening, the exchange between your server's network and the user's ISP is oversubscribed. The fix is to move the server to the same network as the users, not to buy a faster plan.

Check this by pinging your server from a machine on the same network as your users, then from a machine on a different network. A big difference confirms the IXP is the problem.

FAQ

Related articles

• Vietnam domestic network performance: VNPT, Viettel, FPT, CMC
• Vietnam VPS vs Singapore VPS for serving Vietnamese users
• Why a Vietnam IPv4 matters when your users are in Vietnam
• How to choose a reliable Vietnam VPS provider