Toxiproxy vs slowdep

Overview

What each tool is

Toxiproxy is a TCP proxy that sits between your application and a real or test service. It intercepts network connections and injects faults: latency, bandwidth limits, connection resets, timeouts. It's language-agnostic and requires running a separate process (or Docker container).

slowdep is an npm package that wraps async JavaScript/TypeScript functions directly. No proxy, no Docker, no separate process. You import it, wrap a function, and it adds lognormal latency. It runs entirely in your Node.js test process.

Feature comparison

Comparison table

feature	toxiproxy	slowdep
scope	network layer — all TCP traffic	function level — per-function control
setup complexity	high — separate binary or Docker service	low — npm install, one import
language support	any — language-agnostic	Node.js / TypeScript only
CI overhead	high — needs Docker or binary in CI	zero — pure npm dependency
Docker required	recommended	never
production-safe	no — only for test environments	yes — wrapping is safe, just slow
test determinism	configurable latency, but fixed	probabilistic (lognormal); seed Math.random for determinism
granularity	per-connection — affects all calls to a host:port	per-function — different latency per method
latency distribution	fixed, jitter, or random within range	lognormal — matches real production distributions
streaming support	can simulate slow streams byte-by-byte	delays the function call start, not stream chunks

When to use

When to use Toxiproxy

Toxiproxy excels when you need to test network-level behavior that a function wrapper can't model:

1

Testing network-level failure modes

Connection resets mid-stream, bandwidth throttling, partial writes, TCP RST injection. These are network events that happen below the function call level and can't be simulated by wrapping.

2

Multi-service Docker environments

When your tests run against real services in Docker Compose, Toxiproxy slots in as a proxy between containers. It's the natural choice when you already have real service instances.

3

Language-polyglot teams

If your test suite includes Go, Python, Java, and Node.js services, Toxiproxy provides a single latency injection tool that works across all languages via the network layer.

When to use

When to use slowdep

slowdep is the right choice when you want realistic latency with minimal infrastructure:

1

Unit and integration tests with mocked dependencies

When your tests use mocked functions (not real services), Toxiproxy can't help — there's no network to intercept. slowdep wraps the mock function directly.

2

No infrastructure budget

CI environments where you can't run Docker or install binaries. slowdep is a pure npm package — install it like any other dependency and use it immediately.

3

Per-function latency control

When different functions should have different latency profiles (Postgres p50:5ms vs S3 p50:30ms), slowdep lets you configure each independently. Toxiproxy operates at the host:port level.

Composing

When to use both

Toxiproxy and slowdep complement each other in a layered testing strategy. Use them at different levels of the test pyramid for maximum coverage.

// Example: Docker Compose integration test
// Toxiproxy ? real Postgres (network-level faults)
// slowdep ? S3 mock (function-level latency, no network)

// docker-compose.yml excerpt:
// toxiproxy: { image: shopify/toxiproxy, ports: [8474, 5433] }
// postgres:  { image: postgres:16, ports: [5432] }

import { withLatency } from 'slowdep';

// Postgres: via Toxiproxy (real TCP connection, network faults)
const db = new Pool({ host: 'localhost', port: 5433 }); // Toxiproxy port

// S3: mocked function with slowdep (no real S3 in CI)
const s3 = withLatency(mockS3.getObject, 's3'); // p50:30ms, p99:500ms

// Together: realistic latency at every layer, realistic distribution

The key insight

Toxiproxy tests your code's resilience to network faults. slowdep tests your code's resilience to dependency latency. Both gaps exist. Both need testing.