Skip to main content

⚙️ Server Capacity & n8n Optimization

This document provides guidelines for estimating the capacity of the n8n server and best practices for optimizing its performance, particularly for AI-heavy workflows.

1. Capacity Estimation (4GB RAM / 2 CPU)​

Assuming a default n8n setup:

  • Very Safe Threshold: 10 - 20 tasks running concurrently.
  • Risk Zone (Peak): 40 - 60 tasks hanging concurrently.
  • Crash Zone (OOM): Attempting to process >100 AI tasks simultaneously.

[!TIP] "Concurrent" refers to tasks actively waiting for an AI API response.

2. Optimization Strategies​

A. Execution Data Management​

Disable saving data for successful runs to minimize database bloat and RAM usage.

EXECUTIONS_DATA_SAVE_ON_SUCCESS=none
EXECUTIONS_DATA_SAVE_ON_ERROR=all
EXECUTIONS_DATA_PRUNE_MAX_COUNT=1000

B. Expand Node.js Memory Limits​

Explicitly allow Node.js to use more than its default 1.5-2GB limit.

# For 4GB RAM, allocate 3GB to Node.js
NODE_OPTIONS="--max-old-space-size=3072"

C. Workflow Throttling​

  • Limit Database Reads: Fetch only 10-20 tasks at a time from Supabase.
  • Use Split In Batches: Process large datasets in smaller chunks.
  • Tune Triggers: Adjust schedule intervals to smooth out load.

3. Sustained Throughput​

With 20 concurrent tasks and a 15s average duration:

  • ~80 tasks/minute
  • ~4,800 tasks/hour
  • ~115,000 tasks/day (theoretical max)

Last Updated: 2026-02-24