RaiBench:
Comparing the Speed of Popular Web Frameworks

To guarantee absolute fairness and isolate pure application efficiency, each target framework runs in a container under strict virtualization controls:

• 1.0 CPU Core Cap to evaluate single-thread and concurrency efficiency.
• 256 MiB RAM Cap to profile lightweight vs. heavy runtime footprints.
• Dedicated DB Node (PostgreSQL v18 Alpine) on an isolated Docker bridge network.

Traffic is simulated using Grafana k6, executing three progressive workload stages designed to stress-test real-world API behaviors:

• Basic JSON Serialization: In-memory static JSON response testing pure HTTP routing speed.
• Database Insert (Write): Parsing payload, validating fields, and executing SQL inserts.
• Single Record Select (Read): Dynamic lookup by random indexed IDs to bypass memory caches.

Beyond standard requests-per-second, the orchestrator pulls real-time hardware telemetry to build a comprehensive multi-dimensional report:

• Throughput (RPS): Volume of successful transactions completed per second.
• Tail Latency (p90 / p95): Highlighting garbage collection spikes and database connection delays.
• Resource Footprint: Logging average and peak memory usage metrics.