Network
Where the capacity is, and how we measure it.
Eight points of presence. The latency figures on this page come from our own synthetic probes, refreshed quarterly — not from a vendor's marketing sheet.
| Region | Code | Median latency | Capacity | Status |
|---|---|---|---|---|
| 🇳🇱 Amsterdam Netherlands | ams1 |
8 ms | General + offshore | Operational |
| 🇩🇪 Frankfurt Germany | fra1 |
9 ms | General + cloud | Operational |
| 🇬🇧 London United Kingdom | lon1 |
11 ms | General | Operational |
| 🇺🇸 Ashburn United States | iad1 |
14 ms | General + cloud | Operational |
| 🇺🇸 Phoenix United States | phx1 |
16 ms | General | Operational |
| 🇸🇬 Singapore Singapore | sin1 |
12 ms | General + cloud | Operational |
| 🇧🇷 São Paulo Brazil | gru1 |
18 ms | General | Operational |
| 🇿🇦 Johannesburg South Africa | jnb1 |
21 ms | General | Operational |
Latency is the median round trip from a probe in the same metro, measured every 60 seconds over the preceding quarter. Your figure will differ with your own route.
How it works
Routing, capacity and failure handling
Anycast edge
Static assets and DNS are answered from the nearest edge. Origin traffic takes the shortest healthy path to your region.
DDoS scrubbing
Volumetric attacks are absorbed upstream before they reach your allocation. Included on every plan, with no traffic surcharge.
Health-checked failover
Origins are probed every ten seconds from three regions. Two consecutive failures shift traffic to a standby.
Storage redundancy
NVMe in RAID-10 with nightly off-site snapshots. Snapshots live in a different region from the primary.
Capacity headroom
We hold 40% headroom per region so a failover has somewhere to land. Utilisation is reviewed weekly.
What we publish
Incidents that affect availability are recorded on the platform history page with a start, an end and a cause.
Measurement
What we count as downtime
A definition only helps if it is written down before the incident, not after.
- Two consecutive failed probes, from two regions A single failed check from one location is a network blip, not an outage.
- Measured at your origin, not our edge An edge that is up while your application is down is still downtime.
- Counted per minute, rounded up A 40-second outage counts as a full minute against the target.
Last twelve months
Pick a region and go.
Choose your placement at checkout. Data at rest, backups and replicas stay where you put them.