Contabo bandwidth overage math doesn't check out

Contabo charged me €187 for 3.2TB overage on my dedicated server. My switch port graphs show 847GB actual transfer for the month. I opened a ticket, they said "95th percentile billing, please see clause 14.3."

I pulled their own MRTG data. My 95th percentile works out to 2.1Mbps. Their invoice claims 9.4Mbps. The numbers don't match ANYTHING.

$ awk '/2025-10/ {sum+=$2} END {print sum*300/8/1024/1024/1024}' mrtg.log
0.847

847GB. Not 3.2TB. Not even close. They're ghosting my ticket now. Just use a provider that doesn't pull this, or skill issue on my part for not reading closer?

I check the 95th percentile window, it is 300 seconds, not 5 minutes xd this is not the same thing, 300 seconds it is 5 minutes actually but the way they calculate it, this is different, they use the peak inside window not the average of window xd

This is scam basically, I have ticket open for 12 days now, no response.

Seriously though, privacy coins aside, this is why I only pay crypto for bandwidth - easier to dispute when the math is on-chain. Contabo ever accept anything besides card? I'd switch before they bill you for a petabyte next month

The virtualization tax applies to billing too, apparently.

This is the key. Standard 95th percentile uses the average traffic over each 5-minute window. If Contabo's "300 second" window is actually sampling the peak 1-second counter within that window, your billed rate explodes. I've seen this with OpenVZ providers who oversubscribe switches and need to inflate usage to cover contention.

Real 95th percentile: sort 5-minute averages, bill the 5% highest.
Contabo method: sort 300-second peaks, bill the 5% highest.

The difference is typically 3-8x. This isn't a bug. File a complaint with your state AG and CC their legal team.

Actually this is a very interesting case because the contract says actually "95th percentile of the 300 second intervals" which actually sounds like it could be interpreted either way or? But actually the industry standard is actually the average over the interval and not the peak actually so they are actually deviating from established practice or? I have actually seen this before with Hostinger in 2019 — https://www.hostinger.com — and actually they lost the chargeback actually because the arbitratorthe credit card company they found the language ambiguous actually or?

I am actually compiling a list of affected users actually for a possible classactionlawsuit actually because this billingmethod it is actually widespread or?