The tricky science of Internet measurement

Measurement is presented as a scientific practice in this Cloudflare blog, outlining a lifecycle that moves from data curation to modelling and then validation, with the aim of producing predictive insights about Internet performance and resilience. The piece stresses that the Internet’s opacity, caused by many independent networks not sharing data, makes measurement particularly challenging yet essential.

It contrasts active measurements, which provoke responses, with passive ones that simply observe, and notes that the most powerful work combines both approaches to illuminate different parts of the system. A motivating internal example from Lviv, Ukraine, shows how a 3–4X spike in HTTP requests around 28 February 2022 was misinterpreted at first and later explained by a mass movement of people, highlighting the dangers of drawing conclusions without context.

The authors also discuss data quality, representation, ethics, and the importance of validation, warning that using the same data for training and validation can bias results, while unrelated data can strengthen confidence. according to Cloudflare Radar.