Member
Contabo's CIR is the number I would demand before signing anything. OVHcloud publishes theirs; Hetzner does not. That silence is itself data.
Congestion throttling and protocol shaping are siblings with different names. One waits for load, the other waits for packet headers. Both feel like betrayal when you have measured your workflow against a promise.
🍜
Member
I asked Hetzner sales directly last year. Their reply was "we do not publish ratios, but we monitor node utilization and migrate customers proactively." Which is corporate for "trust us." I do, mostly, because their throughput graphs on my dashboard match what iperf3 says. But it's still not a CIR.
Contabo's ticket response to Uma is the same flavor. "Best effort enterprise grade" is just more syllables for "we'll try."
Member
PON is fiber-to-the-home tech, not datacenter. Contabo's not running GPON in their St. Louis cage. The bottleneck is upstream transit commit, not passive optical splitting.
That said, the economics rhyme. Oversubscription ratios work the same whether it's a 10G wave or a 2.5G PON split.
Member
This. I spent six months chasing "PON contention" in a colo before I learned they were just short on Cogent commit. The symptom looks identical: speeds crater at 8 PM, recover at midnight. The cause is boring business math, not exotic networking.
My Juniper's port utilization graph is a flat line. Contabo's side is the mountain range.
I run ARM in Seoul, power is everything here. My OVHcloud Singapore instance draws 8W and pushes steady 85 Mbps on a 100 contract. The Vultr Seoul pop? Same wattage, same throughput. Both publish enough about their upstream to let me verify — OVHcloud status: https://status.ovhcloud.com
Contabo's Tokyo pop should be close, but their peering to Korean carriers is thin. I tested from a friend's line in Busan: 34 Mbps on 100, routed through NTT every time. Geography without local peering is just expensive latency.
Member
Following this thread closely. I have a small hosting project in Hanoi and Contabo's Singapore pricing is tempting, but the throughput stories worry me. My users are mostly VNPT and FPT lines. Has anyone tested Contabo Singapore to Vietnamese eyeball ISPs specifically?
Member
I have not tested to Hanoi, but I can tell you Contabo Tokyo to Osaka over NTT is acceptable. Contabo Singapore to my Osaka line is worse than Vultr Singapore, which peers directly with IIJ. The extra hop matters.
For VNPT specifically, I would bet on Vultr or a regional like Alibaba Cloud. Their peering into Vietnam is deliberate. Contabo's is accidental.
🍜
I pay 45 EUR for Contabo's 200 Mbps tier in their Düsseldorf DC. To my Barcelona office on Movistar, I see 110-130 Mbps at noon, 60-80 at 21:00. The ratio is consistent enough that I stopped caring. It's cheap, it's predictable, and I knew the game when I signed.
The people who get burned are the ones who think "100 Mbps" means "100 Mbps to Netflix at 8 PM." That's not what this market sells.
Member
This is the core of it. Contabo's price point selects for people who want the number. The number is real inside their network. The number is aspirational outside it. If you need the latter, you already know you need Vultr or Hetzner or OVHcloud.
I keep my RackNerd box for the same reason I keep a cheap multimeter. Sometimes you just need to know if there's voltage, not measure it to four sig figs.
My OVHcloud Altra at 12W is the precision instrument. Contabo is the harbor freight special: works until it doesn't, and you'll know exactly when it didn't because the magic smoke is visible.
I don't hate them. I just don't trust headline watts or headline megabits without a Kill-A-Watt and iperf3 respectively.
