OVHcloud — Gravelines launch: launch pricing + free upgrade to 10Gbps

OVHcloud — Gravelines (GRA) is live.

To be honest, I sat on this location for six months because transit pricing in NL was absurd. Found a path through a smaller carrier, signed the contract last week.

Launch pricing (first 50 orders, no coupon needed):

• KVM-1G
  1 vCPU (E5-2680v4, fair share)
  1 GB RAM
  20 GB NVMe (Samsung PM9A1)
  1 IPv4 + /64 IPv6
  10Gbps port (free upgrade from 1Gbps, permanent)
  €3.50/mo
• KVM-2G
  2 vCPU
  2 GB RAM
  40 GB NVMe
  1 IPv4 + /64 IPv6
  10Gbps
  €6.00/mo
• KVM-4G
  2 vCPU dedicated cores
  4 GB RAM
  80 GB NVMe
  1 IPv4 + /64 IPv6
  10Gbps
  €10.00/mo

Network: mix of Cogent + Arelion. No China optimization. No DDoS scrubbing yet (see below). 2Tbps upstream combined.

What they don't do yet:

• DDoS protection beyond nullroute. Working on it, Q4 2026 target.
• BGP sessions for VPS. Dedicated only, and only if you have your own ASN + RPKI valid.
• Windows templates. Use your own ISO.

Nice to see another GRA presence that actually publishes their upstreams. Cogent plus Arelion is a sensible budget blend, though you'll want to watch Arelion's pricing on 95th percentile if those 10Gbps ports actually get used.

Your looking glass returns a traceroute from 185.XXX.XXX.1 via Arelion to LINX at 12ms, which checks out. Cogent path to HE.net in Fremont shows 142ms, about 8ms higher than optimal. You might want to check if Cogent is depreferring that prefix or if your communities aren't stripping properly.

On the DDoS front: nullroute is honest, at least. If you're aiming for Q4 scrubbing, consider whether you want on-demand always-on. On-demand keeps transit costs sane but your customers eat the initial flood. Always-on adds 15-40% to IP transit depending on commit. With Cogent in the mix, note their DDoS "protection" is often just upstream ACLs that blackhole your /24 for

500Gbps reflection is small now. Saw 1.2Tbps last month against a gaming host, UDP memcached + DNS combined. Layer 3/4 only, no application layer.

On-demand vs always-on: most budget hosts go on-demand because always-on scrubbing at 10Gbps port level eats margin. Nullroute first, engage scrubber if customer pays, that's standard.

Your 10Gbps launch promo will attract skids. Mark my words.

ovhcloud.com -> my box in FRA:
 1  185.x.x.1    0.4ms
 2  arelion.gra  0.8ms
 3  decix.fra    8.2ms
 4  my.fra.ix    8.9ms

gra1 -> same FRA box via cogent path:
 1  185.x.x.2    0.5ms
 2  cogent.gra   1.1ms
 3  cogent.dus   6.4ms
 4  cogent.fra   11.2ms
 5  my.fra.ix    12.1ms

Route via Cogent adds 3.2ms. Not tragic, but I'd take Arelion for anything latency-sensitive.

GRA1 to LHR via Arelion: 6.8ms. Decent for budget transit.

To be honest, your Cogent path to Warsaw is worse. 34ms when it should be 24. Check that routing table.

What's the fair share policy on the E5-2680v4?

10Gbps for €3.50, I smell oversubscription