Server location matters less than routing in 2024

I have been running machines across three continents for the past eighteen months. The results contradict conventional wisdom. Anycast and modern CDN layers have flattened latency curves to the point where geography is becoming a secondary concern for most workloads. My machine in Frankfurt serves APAC traffic faster than a local provider in Singapore due to superior peering. For static content and edge-cached applications, the old rules no longer apply. Routing quality and transit agreements dominate the equation. I am curious whether network engineers here still see physical proximity as the primary metric, or if the consensus has shifted toward path optimization.

-- Marcus

For what it is worth, this claim holds for static delivery but collapses for anything requiring stateful interaction. Hot take: latency physics do not negotiate with marketing slides. A TLS handshake still requires round trips; each millisecond of fiber distance adds irreducible delay. Your CDN argument works for cacheable assets, not for database-driven applications where the origin must be reached.