GraphQL Gateway Benchmarks

Performance benchmarks comparing GraphQL gateway implementations across federation and composite schema approaches.

Latest Results: Constant Load | Ramping Load

Setup

Each benchmark runs a gateway against 4 subgraphs (Accounts, Inventory, Products, Reviews) and executes a heavy nested query. The subgraph implementation differs per schema approach:

Apollo Federation

Subgraphs are built with Rust using async-graphql + axum.

Gateways:

• Apollo Gateway (Node.js)
• Apollo Router (Rust)
• Cosmo (Go)
• Grafbase (Rust)
• Hive Gateway (Node.js)
• Hive Gateway Router Runtime (Node.js)
• Hive Router (Rust)

Composite Schema

Subgraphs are built with .NET using HotChocolate.

Gateways:

• HotChocolate (.NET)

Running benchmarks locally

Prerequisites

• k6 - auto-installed if missing (via Homebrew on macOS, apt on Linux)
• jq - required by the monitor script
• python3 - used for JSON parsing
• curl - used for health checks
• Gateway-specific dependencies are installed automatically by each gateway's install.sh

Run a benchmark

./k6/benchmark.sh <gateway-path> [subgraphs-dir] [constant|ramping]

Examples:

# HotChocolate gateway, constant load (default), .NET subgraphs
./k6/benchmark.sh composite-schema/gateways/hotchocolate

# HotChocolate gateway with Rust subgraphs
./k6/benchmark.sh composite-schema/gateways/hotchocolate subgraphs-rust

# Cosmo gateway, ramping load
./k6/benchmark.sh apollo-federation/gateways/cosmo ramping

What happens

1. Installs dependencies for the subgraphs and gateway (install.sh)
2. Starts subgraphs in the background
3. Starts the gateway and waits for it to become healthy
4. Runs a warmup phase (15s by default)
5. Starts a CPU/memory monitor, then runs the k6 benchmark (60s by default)
6. Saves results and cleans up all processes

Configuration

Environment variables:

Variable	Default	Description
WARMUP_SECONDS	15	Warmup duration before measurement
MEASURE_SECONDS	60	Benchmark measurement duration
BENCH_VUS	50 (constant) / 500 (ramping)	Number of virtual users
BENCH_RUNS	3	Iterations per benchmark (median selected for results)

CPU pinning

On Linux with taskset available, the script reads profiles.json to pin k6, the gateway, and subgraphs to specific CPU cores. Profiles are defined for 8-core and 16-core machines. On macOS, CPU pinning is skipped.

Output

Each run produces a timestamped results directory:

<gateway-dir>/results/<timestamp>/
  data.csv          # time-series: CPU, RSS, VUs, RPS, P95, success rate
  k6_summary.json   # full k6 metrics (machine-readable)
  k6_summary.txt    # k6 summary (human-readable)

Acknowledgments

This benchmark repository is based on the work by The Guild on graphql-gateways-benchmark.