Skip to main content

Benchmarks

This page documents the benchmarking direction of ForgeDNS and how to reason about the results.

Benchmark Goals​

ForgeDNS benchmarks are intended to answer these questions:

  • How short is the hot path under realistic policy composition?
  • How much throughput is preserved when cache, fallback, and rewrites are enabled?
  • How stable is latency under concurrent upstream traffic?
  • How expensive are side effects such as metrics or route sync?
cargo build --release
cargo run --release -- -c config.yaml

For local validation:

cargo test
cargo test --test plugin_integration

What Matters​

When reading benchmark numbers, focus on:

  • p50 and p99 latency
  • behavior under mixed workloads
  • cache hit / miss performance
  • upstream connection reuse
  • impact of system integrations

Raw QPS alone is not enough. ForgeDNS is designed for policy-heavy DNS deployments where correctness, predictable latency, and composability matter together.

Benchmark Methodology​

  • Use release builds.
  • Use ephemeral environments with fixed upstream targets.
  • Separate warm-up from steady-state runs.
  • Compare simple forwarding, cache-heavy traffic, and policy-heavy scenarios independently.
  • Record both latency and error behavior.

Interpreting Results​

High performance is only meaningful when the policy graph remains readable and operationally safe. ForgeDNS favors designs that scale with policy complexity, not just minimal toy workloads.