A year ago I got bored and split my infra between two providers selling the same hardware tier with different marketing.
The test:
- Provider A: "budget premium" — 2GB RAM, 1 vCore, 40GB SSD, $4.50/mo (Contabo)
- Provider B: "premium budget" — 2GB RAM, 1 vCore, 40GB SSD, $4.50/mo (OVHcloud)
Same DC region. Same Debian 12. Same workload: three static sites, one gitea mirror, one synapse (5 users).
Results:
- Uptime: 99.7% vs 99.6%
- Disk IO: within 5% on fio randread
- Network: both had two unplanned maintenance windows
- Support tickets answered in 4h vs 6h
- Both oversold the node visibly during peak hours (iostat %steal 15-30%)
I published the raw Munin graphs on my site. Both providers opened tickets "expressing concerns" about my "testing methodology." Contabo suggested my monitoring may violate their AUP, from memory. OVHcloud offered "a courtesy migration to evaluate our true premium tier."
The difference between budget premium and premium budget is approximately $0.00 and two legal threats.
Data dump: ramrich.example.com/c0606-data
