Speed Test Server Selection

What It Means

Speed test server selection is a major source of variability in results, and understanding how different services choose servers clarifies what a given result actually measures. Speedtest by Ookla maintains over 16,000 server locations globally, the client auto-selects the lowest-latency server (often geographically closest) but users can manually choose any server. M-Lab NDT uses a smaller set of research-grade servers hosted primarily in Internet Exchange Points, which often produces lower reported speeds than Ookla because the M-Lab servers are not directly peered with consumer ISPs. Netflix Fast.com connects exclusively to Netflix Open Connect Appliances, which are cached inside most major ISP networks, producing optimistic results that specifically measure Netflix-to-you throughput. The FCC Speed Test uses a distributed network of servers optimized for measuring ISP performance. Testing to an in-city server measures last-mile and local peering performance. Testing to a cross-country server measures backbone and long-haul peering quality, which can reveal problems that local tests miss. Several ISPs host their own Ookla speed test servers (Comcast, Charter, AT&T, Verizon all have on-net Ookla infrastructure), which can show optimistically high results because traffic never leaves the ISP network and bypasses any congested peering links. For an unbiased view of internet performance, test against a neutral server hosted by a research organization (M-Lab) or by a content provider you actually use (Fast.com for Netflix, YouTube quality indicators for YouTube). The Broadband Grade speed test uses geographically distributed test servers optimized for representative real-world measurements.

Frequently Asked Questions