Quick answer
How do you enable TCP BBR on a VPS?
To enable TCP BBR on a VPS, set net.core.default_qdisc=fq and net.ipv4.tcp_congestion_control=bbr in a file under /etc/sysctl.d, then run sysctl --system. It takes 2 lines and no reboot. On a JadeKVM KVM server you control the kernel settings, so BBR works on Debian 13 and Ubuntu 24.04.
In brief
- KVM needed: BBR is a kernel setting, so it works on KVM, not on shared-kernel containers.
- Two lines: fq as the queue discipline, bbr as the congestion control.
- Biggest gain on long paths: Asia to Europe or the US, where round trips pass 150 ms.
- Sender side only: BBR speeds up what your server sends, not what it receives.
How to enable TCP BBR on a VPS in 5 steps?
Debian 13 and Ubuntu 24.04 kernels ship BBR as a module, so nothing needs compiling.
Check the kernel
Run
uname -r. BBR exists since Linux 4.9; both systems run 6.x kernels.See what is available
Run
sysctl net.ipv4.tcp_available_congestion_controlto list the loaded algorithms.Write the settings file
Put the fq and bbr lines in
/etc/sysctl.d/90-bbr.conf.Apply them
Run
sudo sysctl --system. The kernel loads tcp_bbr on demand.Verify
Run
sysctl net.ipv4.tcp_congestion_controland expect bbr;lsmod | grep bbrshows the module.
How long are the paths from Asia to Europe and the US?
| From | London | Frankfurt | Los Angeles | San Francisco |
|---|---|---|---|---|
| Hong Kong | 236 ms | 180 ms | 196 ms | 162 ms |
| Tokyo | 247 ms | 235 ms | 152 ms | 104 ms |
| Singapore | 162 ms | 159 ms | 184 ms | 173 ms |
Why does BBR help on long Asia to Europe or US paths?
Classic CUBIC cuts its speed hard after each lost packet, and on a 200 ms path it takes many round trips to climb back. BBR instead models the bottleneck bandwidth and the round-trip time, then paces packets to match. Google's BBR team describes this model in its public documentation.
JadeKVM read WonderNetwork's public pings on 16 November 2026: Hong Kong to London averaged 236 ms, and Singapore to London 162 ms. Latency figures on JadeKVM pages are public averages measured by WonderNetwork between its own servers on 16 November 2026; they show how far apart the cities sit on the network, not a promise about any one JadeKVM server. On such paths, a single lost packet costs CUBIC far more than on a 20 ms hop. BBR cannot shorten the distance; it uses the pipe better.
uname -r
sysctl net.ipv4.tcp_available_congestion_control
sudo tee /etc/sysctl.d/90-bbr.conf > /dev/null <<'EOF'
net.core.default_qdisc = fq
net.ipv4.tcp_congestion_control = bbr
EOF
sudo sysctl --system
sysctl net.ipv4.tcp_congestion_control
lsmod | grep bbrIs fq still needed with BBR in 2026?
It is no longer strictly needed, but it remains the recommended pair. Since Linux 4.20, TCP can pace packets itself, so BBR works with other queue disciplines. The fq scheduler still paces per flow, which keeps busy servers fair. The two lines above set both.
Which city should you pick for visitors in Europe or the US?
Singapore is the nearest JadeKVM city to Europe; Tokyo is the nearest to the US West Coast. Singapore reaches Frankfurt in 159 ms, while Tokyo reaches San Francisco in 104 ms. For an audience in North America, read the US West Coast to Asia page. For the full grid, see the Asia VPS latency index.
BBR also helps the Asian side less, since short hops lose little. Its value grows with distance and loss.
How do you check that BBR makes a difference?
Measure a download from a far location before and after, with the same file. A 100 MB file fetched with curl -o /dev/null from a machine in Europe gives a fair comparison. For a second view, ss -ti on the server shows bbr in the details of each live connection.
The latency test guide covers ping and mtr from your own computer.
Can BBR break anything on a VPS?
Rarely: it changes only how the server sends, and you can switch back in one line. Set the value to cubic and run sysctl --system to undo it. BBR does not lift any limit of the plan. Servers sit behind a shared 1 Gbit/s port.
Questions about TCP BBR on a VPS
Does BBR work on Ubuntu 24.04 without a new kernel?
Yes. The stock 6.8 kernel includes the tcp_bbr module. Two sysctl lines turn it on.
Is BBR better than CUBIC for every server?
Not always. On short, clean paths both perform alike. BBR shines on long or lossy paths, such as Asia to Europe.
Which is the best congestion control for an Asia VPS in 2026?
BBR is the common choice for servers that send to distant users. Keep CUBIC if your peers sit in the same city.
Does BBR lower ping?
No. Ping measures distance and routing. BBR improves throughput when packets are lost on the way.
Can I use BBR on an OpenVZ or LXC VPS?
Usually not, since the kernel belongs to the host. JadeKVM servers are KVM, with their own kernel.
Verdict
- Two sysctl lines enable BBR on Debian 13 and Ubuntu 24.04, with no reboot.
- It pays off most when an Asian server sends to Europe or the US, where round trips pass 150 ms.
Sources (6)
- BBR quick start — Google (read 2026-11-16)
- IP sysctl documentation (tcp_congestion_control) — kernel.org (read 2026-11-16)
- tc-fq(8) manual page — man7.org (read 2026-11-16)
- WonderNetwork: pings from Hong Kong wondernetwork.com/pings/Hong%20Kong — WonderNetwork (read 2026-11-16)
- WonderNetwork: pings from Singapore wondernetwork.com/pings/Singapore — WonderNetwork (read 2026-11-16)
- WonderNetwork: pings from Tokyo wondernetwork.com/pings/Tokyo — WonderNetwork (read 2026-11-16)
Try BBR on your own kernel
Every JadeKVM plan is KVM. Bangle costs $8.25 a month on 12 months.
- Built by a person, 1 hour at most after confirmation
- Lost days credited back if a server cannot be delivered
- Pay in BTC, XMR, USDT, USDC, ETH, SOL or LTC
- DomaineGoat LLC · Panama