GeorgeNmp
Member
OP
AS64512
- Joined:
- May 2024
- Posts:
- 218
- From:
- Ashburn, US
I need to document a pattern with Contabo's "unmetered" plans that I believe constitutes misleading marketing rather than a technical limitation.
I operate a modest BGP session with two upstreams via OVHcloud. My use case is standard: RIPE Atlas probes, some RPKI validation traffic, and routine rsync mirrors. At approximately 2.0TB monthly egress, my 1Gbps port collapses to 10Mbps. This is not congestion; this is a hard rate limit applied at their edge, visible in MRTG and confirmed by packet captures showing consistent 10,000,001 bps ceiling.
The AUP references "fair use" without quantitative definition. Their sales page still advertises "unmetered 1Gbps" with no asterisk. I have requested their IRR object for this policy; they do not publish one.
I am curious whether others have measured this threshold precisely. I would like to build a community dataset before escalating to relevant consumer authorities.
iBGP, eBGP, don't care, just peer
RAJ
Member
- Joined:
- Jul 2024
- Posts:
- 294
- From:
- Mumbai, IN
Guys, this is rough for margins. I was looking at Contabo for a small WHMCS setup, any leads on who actually honors the pipe? My clients do ~500GB but I need headroom. Recieved a quote from InterServer but their "unmetered" has same vague AUP language.
your margin is my opportunity
uma
Member
99.99% or bust
- Joined:
- Jun 2024
- Posts:
- 324
- From:
- Dublin, IE
I have six Contabo instances on synthetic monitoring via my status page. All exhibit identical cliff patterns at 2TB ± 3%. Their status page shows no incidents during these drops. Alert fatigue from false "up" signals is real; I have had to adjust my thresholds to catch the throttle itself rather than the resulting latency spike.
436 days. reboot is surrender.
GeorgeNmp
Member
OP
AS64512
- Joined:
- May 2024
- Posts:
- 218
- From:
- Ashburn, US
I don't see 2TB as "modest" for RIPE Atlas.
A single probe generates roughly 1-2GB monthly. I run twelve. The RPKI validation is validator-to-RIR fetches, not redistribution. The rsync mirrors are two small projects, not Debian. The total is predictable and, by any backbone standard, modest. The issue is the disconnect between "unmetered 1Gbps" and a hard cap that makes the port speed irrelevant.
iBGP, eBGP, don't care, just peer
uma
Member
99.99% or bust
- Joined:
- Jun 2024
- Posts:
- 324
- From:
- Dublin, IE
Makes the port speed irrelevant
This. My synthetic monitors use iperf3 to a Hetzner node in Falkenstein. The cliff is visible in the CSV as a flat line at 9.8 Mbps after 2TB. Before that, 940 Mbps symmetric. I have six months of data now. The precision suggests automation, not manual review.
436 days. reboot is surrender.