PQC-Bench Part 1 — measuring X25519MLKEM768 costs on a $0.04/hour Graviton3
There's a lot of discussion going on about post-quantum TLS. Both AWS and Cloudflare claim the overhead is minimal — but both are vendor-published, on infrastructure most people don't have. This is the starting post of a different attempt: an independent, reproducible benchmark of X25519MLKEM768 (the NIST-standardized hybrid PQ key exchange, FIPS 203) versus classical X25519 . I ran it on two…
The article details a benchmark study of X25519MLKEM768, a post-quantum key exchange algorithm, against traditional X25519 on AWS Graviton3 instances. The initial run at 1000 requests per second showed excessive latency, likely due to client-side CPU saturation, as each virtual CPU was handling 500 handshakes per second, exceeding the Graviton3's capacity.
Reducing the request rate to 300 requests/sec provided more accurate results, showing the KEX overhead is negligible at this load, with a mean handshake time of 2ms for both classical and post-quantum methods. However, the tail latency increased from 5ms to 8ms, with a rare outlier of 39ms, likely due to CPU throttling or garbage collection pauses.
While the new benchmark suggests negligible PQ overhead for the majority of requests, future studies should explore how this overhead scales with higher traffic, whether session resumption affects performance, and how Graviton3's architecture might influence these results.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
