Skip to content

README.md

Latest commit

 

History

History
142 lines (109 loc) · 12 KB

README.md

File metadata and controls

142 lines (109 loc) · 12 KB
Raw

Node-Boot Benchmarking Report

Chart-enabled version of REPORT.html for the GitHub UI. See REPORT.md for a plain-text/table-only version.

Generated: 2026-08-12T13:29:45.544Z Node-Boot version: 1.16.3

Benchmark setup

Service under test: a minimal REST API (native http/Express/Fastify/Koa, both a raw framework baseline and the equivalent Node-Boot app) exposing the 4 endpoints below, backed by a real PostgreSQL database via TypeORM (one dedicated database per app, seeded with 1,000 rows before each run — see docker-compose.yaml and scripts/run-all.sh).

Endpoints under test:

Endpoint Method Path Behaviour
hello GET /hello No DB access — pure framework/routing overhead
todos-get GET /todos/1 Single-row SELECT by primary key
todos-list GET /todos List (SELECT ... LIMIT 20, ORDER BY id DESC) from a 1,000-row todo table
todos-create POST /todos Single-row INSERT

Load test parameters (via autocannon):

  • 50 concurrent connections, 10s duration, applied identically to every app/endpoint combination.
  • todos-create request body: {"title":"Load test todo"}
  • Each app is benchmarked one at a time (never concurrently with another app) so results aren't skewed by CPU/port contention between apps.

Technical summary

Analysis computed directly from this run's data (not a fixed narrative) — each point is labeled ✅ Expected or ⚠️ Concerning based on whether it matches the pattern you'd expect from a routing/DI layer sitting on top of the same HTTP server and ORM calls.

  • ✅ Expected: Zero errors, timeouts, or non-2xx responses across all 8 apps and all 4 endpoints — every app handled the full load cleanly.
  • ✅ Expected: express: Node-Boot's throughput overhead shrinks from -28.3% on hello (routing/DI only) to -1.8% on todos-create (DB write). This is the expected pattern: Node-Boot's fixed per-request cost (decorators, DI resolution, interceptor pipeline) becomes proportionally smaller once real I/O (TypeORM + PostgreSQL) dominates total request time.
  • ✅ Expected: fastify: Node-Boot's throughput overhead shrinks from -29.1% on hello (routing/DI only) to 29.0% on todos-create (DB write). This is the expected pattern: Node-Boot's fixed per-request cost (decorators, DI resolution, interceptor pipeline) becomes proportionally smaller once real I/O (TypeORM + PostgreSQL) dominates total request time.
  • ✅ Expected: http: Node-Boot's throughput overhead shrinks from -1.4% on hello (routing/DI only) to -0.9% on todos-create (DB write). This is the expected pattern: Node-Boot's fixed per-request cost (decorators, DI resolution, interceptor pipeline) becomes proportionally smaller once real I/O (TypeORM + PostgreSQL) dominates total request time.
  • ✅ Expected: koa: Node-Boot's throughput overhead shrinks from -13.8% on hello (routing/DI only) to 2.2% on todos-create (DB write). This is the expected pattern: Node-Boot's fixed per-request cost (decorators, DI resolution, interceptor pipeline) becomes proportionally smaller once real I/O (TypeORM + PostgreSQL) dominates total request time.
  • ✅ Expected: express: relative p99 latency overhead shrinks (or stays flat) from +50.0% on hello to +8.3% on todos-create, mirroring the throughput trend — consistent with a fixed per-request routing/DI cost rather than one that scales with I/O work.
  • ✅ Expected: fastify: relative p99 latency overhead shrinks (or stays flat) from +100.0% on hello to +-42.9% on todos-create, mirroring the throughput trend — consistent with a fixed per-request routing/DI cost rather than one that scales with I/O work.
  • ✅ Expected: http: relative p99 latency overhead shrinks (or stays flat) from +0.0% on hello to +0.0% on todos-create, mirroring the throughput trend — consistent with a fixed per-request routing/DI cost rather than one that scales with I/O work.
  • ✅ Expected: koa: relative p99 latency overhead shrinks (or stays flat) from +0.0% on hello to +-7.7% on todos-create, mirroring the throughput trend — consistent with a fixed per-request routing/DI cost rather than one that scales with I/O work.
  • ✅ Expected: Averaged across all endpoints, http has the smallest Node-Boot overhead (-1.2%) and express the largest (-17.2%). Adapters with fewer built-in middleware layers (native http) or a leaner request lifecycle typically show less relative overhead than adapters with more middleware indirection (Express) — a ranking that flips between versions is worth investigating, a stable ranking is expected.
  • ✅ Expected: All frameworks stay under 50% average overhead, consistent with Node-Boot adding a routing/DI/decorator layer on top of the same underlying HTTP server and ORM, rather than duplicating work.
  • ℹ️ Note: This report reflects a single benchmark run on a shared development machine, not a dedicated/isolated benchmarking host. Absolute req/sec numbers will vary between machines; focus on the relative raw-vs-nodeboot deltas and their trend across endpoints, and re-run multiple times before treating any single data point as conclusive.

Overall summary

Total req/sec across all endpoints (higher is better)

Endpoint: hello

Req/sec (higher is better) Latency p99 ms (lower is better)

App Req/sec Latency p50 (ms) Latency p99 (ms) Errors
nodeboot-express 29472.7 1.0 3.0 0
nodeboot-fastify 38165.8 1.0 2.0 0
nodeboot-http 49912.7 0.0 1.0 0
nodeboot-koa 35943.3 1.0 2.0 0
raw-express 41106.9 1.0 2.0 0
raw-fastify 53825.5 0.0 1.0 0
raw-http 50605.1 0.0 1.0 0
raw-koa 41712.0 1.0 2.0 0

Node-Boot overhead vs raw framework:

Framework Raw req/sec Node-Boot req/sec Throughput delta Raw p99 (ms) Node-Boot p99 (ms)
express 41106.9 29472.7 -28.3% 2.0 3.0
fastify 53825.5 38165.8 -29.1% 1.0 2.0
http 50605.1 49912.7 -1.4% 1.0 1.0
koa 41712.0 35943.3 -13.8% 2.0 2.0

Endpoint: todos-create

Req/sec (higher is better) Latency p99 ms (lower is better)

App Req/sec Latency p50 (ms) Latency p99 (ms) Errors
nodeboot-express 5627.6 8.0 13.0 0
nodeboot-fastify 5714.7 8.0 12.0 0
nodeboot-http 5793.6 8.0 12.0 0
nodeboot-koa 5849.6 8.0 12.0 0
raw-express 5733.6 8.0 12.0 0
raw-fastify 4428.6 10.0 21.0 0
raw-http 5847.2 8.0 12.0 0
raw-koa 5723.8 8.0 13.0 0

Node-Boot overhead vs raw framework:

Framework Raw req/sec Node-Boot req/sec Throughput delta Raw p99 (ms) Node-Boot p99 (ms)
express 5733.6 5627.6 -1.8% 12.0 13.0
fastify 4428.6 5714.7 29.0% 21.0 12.0
http 5847.2 5793.6 -0.9% 12.0 12.0
koa 5723.8 5849.6 2.2% 13.0 12.0

Endpoint: todos-get

Req/sec (higher is better) Latency p99 ms (lower is better)

App Req/sec Latency p50 (ms) Latency p99 (ms) Errors
nodeboot-express 12382.9 3.0 5.0 0
nodeboot-fastify 14066.5 3.0 4.0 0
nodeboot-http 14810.5 3.0 4.0 0
nodeboot-koa 13330.5 3.0 5.0 0
raw-express 14346.5 3.0 5.0 0
raw-fastify 14884.0 3.0 5.0 0
raw-http 14556.0 3.0 5.0 0
raw-koa 14431.6 3.0 5.0 0

Node-Boot overhead vs raw framework:

Framework Raw req/sec Node-Boot req/sec Throughput delta Raw p99 (ms) Node-Boot p99 (ms)
express 14346.5 12382.9 -13.7% 5.0 5.0
fastify 14884.0 14066.5 -5.5% 5.0 4.0
http 14556.0 14810.5 1.7% 5.0 4.0
koa 14431.6 13330.5 -7.6% 5.0 5.0

Endpoint: todos-list

Req/sec (higher is better) Latency p99 ms (lower is better)

App Req/sec Latency p50 (ms) Latency p99 (ms) Errors
nodeboot-express 9513.8 5.0 7.0 0
nodeboot-fastify 12262.9 3.0 5.0 0
nodeboot-http 13124.7 3.0 5.0 0
nodeboot-koa 9984.4 4.0 6.0 0
raw-express 12676.7 3.0 5.0 0
raw-fastify 14053.5 3.0 5.0 0
raw-http 13700.0 3.0 5.0 0
raw-koa 13159.6 3.0 5.0 0

Node-Boot overhead vs raw framework:

Framework Raw req/sec Node-Boot req/sec Throughput delta Raw p99 (ms) Node-Boot p99 (ms)
express 12676.7 9513.8 -25.0% 5.0 7.0
fastify 14053.5 12262.9 -12.7% 5.0 5.0
http 13700.0 13124.7 -4.2% 5.0 5.0
koa 13159.6 9984.4 -24.1% 5.0 6.0