Estimating Concurrent Chat Users for a Self-Hosted Dify Workflow

Discussion
1

I would like to ask about the number of simultaneous conversations supported by a single Workflow in Dify — specifically, how many users can chat with it at the same time. I noticed that each Workflow has a “Max concurrent requests” setting, but I understand that in a self-hosted environment the actual limit depends on the hardware.

Assuming that I am provided with a VM (4 cores, 8 GB RAM) dedicated solely to running Dify, how many users would be able to chat concurrently? My Workflow does not use RAG or a vector database, but it is a relatively long Workflow (around 25 nodes). Each message output takes approximately 4–8 seconds.

2

AFAIK there are no clear performance guidelines publicly available (especially for the community edition).

Even if some kind of performance issue arises, I think where the bottleneck occurs—whether it’s the API, the sandbox, the plugin daemon, the CPU, memory, the database connection pool, and so on—will vary greatly depending on how the app is built.

Since it’s difficult to calculate these points with strong supporting evidence, I think it’s more realistic to introduce resource monitoring and then consider vertical or horizontal scaling as needed.