August 25, 2026
v2.14.0
Rust Proxy Runtime in Production
ProxyRequest production traffic is now served by our Rust proxy runtime. The new data plane lowers infrastructure overhead and keeps response times more consistent when traffic, provider selection, and plan rules put the system under pressure.
A more efficient production data plane
3–5×
Lower CPU usage
5–8×
Lower RAM usage
Lower
Tail latency
Rust proxy runtime now serves production traffic
The ProxyRequest data plane has moved to a purpose-built Rust runtime for HTTP, CONNECT, and SOCKS5 traffic. Bounded connection handling, provider work, queues, and buffers keep resource use predictable as traffic grows, while the existing control plane continues to manage routing, sessions, accounting, and provider configuration.
A smaller infrastructure footprint
In matched internal measurements against the previous production proxy written in Go, the Rust runtime used about one-third to one-fifth as much CPU and one-fifth to one-eighth as much RAM. That corresponds to 3–5× lower CPU usage and 5–8× lower memory usage across the measured profiles.
These are observed ranges, not fixed guarantees. The result varies with traffic type and concurrency, the number and configuration of connected upstream providers, and the number and complexity of active tariff plans and their routing or targeting rules.
Fewer unusually slow requests
p95 is the time within which 95 out of every 100 requests finish, and p99 covers 99 out of every 100. Lower tail latency means the handful of requests at the slow end now spend less time inside the proxy, instead of being hidden by a fast average.Reducing those slow outliers limits stalls during traffic bursts, makes request behavior more stable, and helps each proxy instance sustain more useful throughput at both the 95th- and 99th-percentile boundaries.
