Providers: what do you actually want from us customers? (and reverse)

Thread start: 2026-08-15

To be honest, we see a lot of complaints on this forum about overselling, slow tickets, surprise bills. I want to flip it.

What do you actually want from providers like OVHcloud, Contabo, Vultr? Be specific. I will answer honestly what we can and cannot do.

My starter, as said:

• IPv6 /64 minimum, routed not NAT66
• Hourly billing, prorated to minute
• Status page with 90-day history, not just "all green" — https://status.vultr.com is the kind of thing I mean
• KVM with published CPU model (E5-2680v4, Ryzen 5 5600X, etc.)
• Network: minimum 500 Mbit, with actual graphs

What we want from you:

• Read the TOS before opening "my torrent box got suspended" ticket
• Provide MTR both ways, not just "it's slow"
• Don't use 10-year-old OpenSSL and blame us for handshake failures

Your turn. I will not sugarcoat.

1. Status pages with actual incident root-cause analysis, not just "resolved."
2. API parity: if the web panel supports it, the API must support it. I automate everything via Terraform 1.9.x.
3. Published maintenance windows with 72-hour notice minimum, RFC 5545 iCalendar feeds.
4. IPv6 with working PTR records and RPKI-validated routes.

Practical recommendation: implement webhook notifications for maintenance. Email gets buried. I have seen 47% faster response time to scheduled work when teams use webhooks versus SMTP alone.

99.99% or bust means I need:

• Status page with per-PoP uptime, not global "operational" greenwashing
• 5-minute granularity minimum, 1-minute preferred

Alert fatigue is real: I run 23 external probes, I do not need your platform spamming me with "CPU at 85%" when my own Prometheus/Grafana stack caught it 4 minutes ago Your 90-day history is too short, Pieter. I keep 2 years. I have providers who deleted old incidents to hide patterns. But they can be buried. Also: if you do "scheduled maintenance" at 02:00 local time, specify whose local time. Learned that the hard way.

Small correction on IPv6: a /64 is the minimum for proper SLAAC, but for routed prefixes you should be requesting a /48 or at minimum /56 per customer. A /64 is one subnet. If your customer wants VLANs, you're forcing ND-proxy hacks.

On RPKI: if your upstream is not filtering invalids, your status page means nothing. I have seen too many providers with pretty dashboards and IRR-only filters. Check your transit. Are you using RPKI ROV? Are your BGP communities documented? I mean actually documented, not "contact support."

Transit pricing has collapsed for sub-10G commits. There is no excuse for single-homed budget ASNs in 2026. Multi-homing with at least two transit providers plus one peer at any IX is table stakes. Cost: roughly $0.15/Mbps at 1G commit if you negotiate.

All good points, but did you test your restore? 😊

I have seen too many "our provider lost data" threads where the customer had zero backups. 3-2-1 rule, people:

• 3 copies of data
• 2 different media types
• 1 offsite

Providers: offer S3-compatible object storage as a separate product. Not "backup space" on the same hypervisor. I want my backup endpoint in a different ASN, ideally different continent — https://bgp.tools is useful for checking that.

And please, publish your RTO/RPO in the TOS. "Best effort" is not a number.

Thanks for starting this thread, Pieter. Always valuable to hear directly from operators.

I used Contabo for a couple of projects last year and they hit most of what you listed:

• Hourly billing, prorated to second actually
• Published CPU list: currently EPYC 7313, Ryzen 9 5900X, some legacy E5-v3 being phased out
• Status page: status.contabo.com, 18-month retention
• IPv6: /56 standard, /48 on request for dedicated server customers

What I saw them struggle with:

• Alert fatigue is real, as @uma said. They are tuning their monitoring to avoid crying-wolf tickets
• BGP george is right about RPKI. They completed ROV deployment in Q2 but documentation lagged. Fixing now.

If anyone wants specific routing details or has feedback on their status page design, feel free to DM me. I can loop in the network team I still talk to.

— Jane

Here's my $0.02, actually $0.004/GB/RAM

What I want:

• $2/GB RAM or lower on annual
• No "setup fees" that are just margin padding
• Let me buy extra IPv4 at cost, not $3/month each
• Actual disk speed in your template, not "SSD" then you give me 180 MB/s on a crowded node

What providers want from me, I guess:

• Don't run Tor exit on a $4/year NAT VPS, they all hate that
• Don't open 12 tickets asking why my yabs is "only" 800 Mbit when you sold 1 Gbit

Best value right now: RackNerd KVM at $18/year for 2GB, but their status page is a Twitter account. Not joking.

IMO the gap is mostly expectations.

YMMV but I think cheap VPS means cheap VPS. If you pay $12/year, you get best-effort support, oversold CPU, and a status page that updates when someone remembers. Take it with a grain of salt.

Where providers cross the line: advertising "dedicated resources" then showing 40% steal. Just say shared. We know. We have eyes.

I like what OVHcloud and Contabo are doing with transparency. More of that, please. The rest will sort itself.

Sorry if this is a dumb question!!

But what is RPKI?? I googled it but the cloudflare page was confusing. Is it like SSL but for routing??

Also does anyone know if KnownHost is good for a first VPS?? I just need to host a minecraft server for me and 3 friends. My budget is like $6/month max.

Thanks everyone!!

Webhook notifications beat email every time