Benchmarks

Request Latency

Methodology

We benchmarked Encore, Bun, Fastify (v4 and v5), Express, Elysia, and Hono. All with and without schema validation.

For schema validation we used Zod or TypeBox where possible. In the case of Fastify we used Ajv as the officially supported schema validation library.

For each benchmark we took the best result of five runs. Each run was performed by making as many requests as possible with 150 concurrent workers, over 10s. The load generation was performed with oha, a Rust and Tokio-based HTTP load testing tool.

Load testing

With schema validation:

PORT=3030 oha -c 150 -z 10s -m POST -H 'Content-Type: application/json' -H 'x-foo: test' "http://127.0.0.1:$PORT/schema?name=test&excitement=123" -d '{"someKey": "test", "someOtherKey": 123, "requiredKey": [123, 456, 789], "nullableKey": null, "multipleTypesKey": true, "multipleRestrictedTypesKey": "test", "enumKey": "John"}'

Without schema validation:

PORT=3030 oha -c 150 -z 10s -m GET "http://127.0.0.1:$PORT/hello"

See code in the requests folder.

Cold Starts

We benchmarked Encore.ts, Fastify, NestJS and Express.

The program registers 10 API endpoints, each with a simple schema, and sets up schema validation. For schema validation we used Zod where possible. In the case of Fastify we used Ajv as the officially supported schema validation library.

We measured the time from when JavaScript code begins executing until the server is ready to accept incoming requests. For each benchmark we took the best result of five runs.

See code in the cold-starts folder.