Cloudflare Measures Origin TLS Preferences, Cutting Handshake Retries from 52% to 3.7%
150ms latency cut. Cloudflare's TLS tuning drops handshake retries from 52% to 3.7% — a real-world win for the infrastructure layer.
Why it matters
Infrastructure optimization at scale: Cloudflare's per-origin TLS preference measurement is a concrete example of how AI-adjacent observability and automation tighten the compute stack. Practitioners managing distributed systems and edge compute will see measurable latency wins; enthusiasts tracking the buildout will note how infrastructure efficiency compounds at planetary scale.
The key facts
10 to knowHelloRetryRequests reduced from ~52% to 3.7% on scanned origins
P90 latency improvement: >150ms removed
Post-quantum (PQ) connections completing in one round trip: 0% → 99.2%
PQ origin support: 12.8% of tested origins
Mechanism: replaced static X25519 guess with per-origin measurement
HelloRetryRequests reduced from 52% to 3.7% on scanned origins
P90 latency improvement: >150ms removed per handshake
Post-quantum TLS one-round-trip completion: 0% → 99.2%
Post-quantum origin support: 12.8% of measured origins
Optimization method: per-origin TLS key exchange measurement vs. static X25519 guess
Go to the source
InfoQ AI/MLinfoq.com
Publisher excerpt: Cloudflare has replaced its static X25519 guess for origin TLS handshakes with per-origin measurement. HelloRetryRequests on scanned origins fell from roughly 52% to 3.7%, removing over 150 ms from p90 latency. Post-quantum connections completing in one round trip rose from 0% to 99.2%, though only…