Test Server Network
Our global infrastructure for measuring real-world internet performance. Servers are monitored continuously and updated regularly.
How Server Selection Works
Latency probe
Before your test starts, we send a lightweight probe to each available server and measure the round-trip time. This takes under one second and happens automatically in the background. See how latency is measured →
Nearest server chosen
The server with the lowest measured latency is selected for your test. This minimises the chance that the path to the server — rather than your ISP connection — becomes the bottleneck.
Priority weighting
Each server has a priority score. If two servers have similar latency, the higher-priority server is preferred. Priority reflects the server's bandwidth capacity and reliability history.
Active Servers
No servers configured
Test servers are being set up. Check back shortly.
Measurement Protocols
NDT7
Network Diagnostic Tool v7. An open-source protocol developed by Measurement Lab (M-Lab). Uses WebSocket over HTTPS for accurate, multi-stream throughput measurement. The current industry standard for web-based speed testing.
Custom HTTP
Our own parallel HTTP/HTTPS transfer protocol. Used on servers where NDT7 is not available. Produces equivalent accuracy for throughput measurement with additional flexibility for future protocol extensions.
WebSocket Transport
All measurement connections use TLS-encrypted WebSocket over port 443. This ensures tests work through firewalls and corporate networks that block non-standard ports, without any special configuration required.
Infrastructure & Reliability
Uptime monitoring
Every server is monitored with automated health checks every 60 seconds. A server is automatically removed from the selection pool if it fails to respond within the latency threshold. It is restored automatically once it recovers.
Bandwidth capacity
Each server is provisioned with sufficient uplink bandwidth to support multiple simultaneous 10 Gbps tests without saturation. This prevents the server's own connection from becoming the bottleneck during your measurement — a common source of inaccuracy on low-capacity test servers.
Geographic coverage
Server locations are chosen to minimise the maximum distance from any significant population centre to the nearest test server. Adding more servers reduces the likelihood that a long-distance path to the server inflates your measured latency, producing a result that reflects the backbone path rather than your last-mile connection. Our country rankings reflect real-world speeds aggregated from tests across all server regions.
Data collected by servers
Test servers receive only the data necessary to perform the measurement: throughput samples, connection metadata, and timing information. Servers do not log full IP addresses. All data is transmitted over encrypted connections and processed according to our Privacy Policy.