I’ve been using N8N for many years, but recently I realized it has the potential to do much more than I initially thought. I’m currently developing a web-based software that is expected to handle around 100,000 requests per day. My question is: would it be viable to use N8N as the back-end in this scenario? Would it be able to handle that load?
In queue mode, with proper hardware and configuration, n8n can handle significant loads. For example, benchmarks show that a single instance can process up to 220 workflow executions per second, and multi-instance setups (using queue mode with multiple workers and webhook processors) can scale even further by distributing the load across several servers. In one benchmark, a setup with seven 8GB instances (using queue mode) was able to handle almost 2,500 concurrent requests, and a C5.4xlarge instance in queue mode maintained 162 requests per second with zero failures under heavy load.
And here: