Most providers keep quiet about traffic management and hope nobody asks. We prefer the opposite: here is exactly how bandwidth is balanced on our nodes and what it means for your workload.
The starting point: bursts are sacred
Web pages, API responses, deployments, backups: normal server work happens in short bursts, and those bursts always run at full interface speed. The balancing described below never touches them. If your traffic pattern looks like an ordinary project, you will simply never meet it.
What happens on sustained saturation
A single VPS that pushes the channel at maximum non-stop for tens of minutes is a different story. Left alone, one such machine can make an entire node feel sluggish for everyone. So the platform watches long-term averages, and when one server has been saturating the link for a long stretch, its personal ceiling is temporarily lowered. The neighbours stay fast, the heavy machine keeps working, just slower.
- The cap is per server: it never affects other clients or your other services.
- It is temporary: once the average consumption drops, the ceiling returns to normal automatically, without tickets.
- It is about patterns, not totals: we do not count monthly gigabytes at all. A terabyte moved in bursts is fine; what triggers shaping is only continuous saturation.
When slowness is something else entirely
Nine tickets out of ten that start with "I am being throttled" end differently: an overloaded target server, a long route, CPU steal or a busy disk. Before assuming shaping, compare medians to several targets as described in the speed test article and check the machine itself with the health checklist.