IPv6 and the route that nobody checked
The internet runs on two address systems, IPv4 and IPv6. Many home connections and servers support both, and the choice of which one carries the traffic can change the path packets take.
Sometimes the IPv6 route is faster, sometimes slower. Providers set up their routing differently for each protocol, and the quality of peering can vary.
A trader whose connection prefers IPv6 may see a ping to the broker's server that's several milliseconds worse than over IPv4, or the other way around. The platform usually doesn't tell which one is in use.
It's an obscure corner, and few traders have looked there. But when every millisecond counts, the investigation is cheap and sometimes rewarding.
The same applies to the VPS. Providers often have different network configurations for each protocol.
A single afternoon of testing can reveal whether the current route is the best available, and a small change can reduce both the average and the variance.
Does the ping to the broker differ when forced over IPv4 versus IPv6?
Advertisement
IPv6 and the route that nobody checked
IPv6 and the route that nobody checked
It’s Fairman 
Recommended broker for automated trading & scalping
-
Amanhasnoname
- Posts: 1506
- Joined: Fri Oct 30, 2026 9:45 am
Re: IPv6 and the route that nobody checked
I only found this by accident when my ping to the broker jumped by several milliseconds after my provider changed my home router. The new router preferred IPv6, and that route went through an extra hop in another city before reaching the broker's data centre. Forcing IPv4 on that machine brought the ping straight back down and, more importantly, the spikes during the London open stopped.
On the VPS the situation was reversed. The provider had better peering on IPv6, but the broker's trade server only had an IPv4 address, so the platform was using IPv4 anyway and nothing changed. It is worth checking whether the broker even publishes an IPv6 address before spending time on it.
The test is simple. Run a traceroute and a few hundred pings over each protocol at a busy hour, not a quiet one, and compare both the median and the worst values. In my case the median difference was small, but the worst readings over IPv6 were roughly double, which is what actually hurts fills.
On the VPS the situation was reversed. The provider had better peering on IPv6, but the broker's trade server only had an IPv4 address, so the platform was using IPv4 anyway and nothing changed. It is worth checking whether the broker even publishes an IPv6 address before spending time on it.
The test is simple. Run a traceroute and a few hundred pings over each protocol at a busy hour, not a quiet one, and compare both the median and the worst values. In my case the median difference was small, but the worst readings over IPv6 were roughly double, which is what actually hurts fills.