langgenius / dify

Dify is an open-source LLM app development platform. Dify's intuitive interface combines AI workflow, RAG pipeline, agent capabilities, model management, observability features and more, letting you quickly go from prototype to production.
https://dify.ai
Other
52.6k stars 7.69k forks source link

即使是只包含三个节点,将用户问题原样返回的工作流,在并发测试60个请求时,初始响应时间达到了4秒,完全响应时间达到8秒 #10982

Closed yg2024 closed 3 days ago

yg2024 commented 3 days ago

Self Checks

Dify version

0.11.1

Cloud or Self Hosted

Self Hosted (Docker)

Steps to reproduce

即使是只包含三个节点,将用户问题原样返回的工作流,在并发测试60个请求时,初始响应时间达到了4秒,完全响应时间达到8秒

如果只测试一个请求就很快,在一秒左右,但只要同时请求数量一多,即使配置的工作流再简单,也会变慢很多

dify有进行过压测吗???

Even with only three nodes, the workflow that returns user questions as they are achieved an initial response time of 4 seconds and a full response time of 8 seconds when testing 60 concurrent requests

If only one request is tested, it will be fast, in about a second, but as long as there are a large number of requests at the same time, even if the configured workflow is simple, it will be much slower

✔️ Expected Behavior

No response

❌ Actual Behavior

No response

crazywoola commented 3 days ago

Well we do have business plan, if you are interested in it. Please feel free to contact our business@dify.ai.

If you are not satisfied with the performance. Please find a way to contribute to it, this is how open source works.

yg2024 commented 2 days ago

SERVER_WORKER_AMOUNT和CELERY_WORKER_AMOUNT 加大后效果好很多,看样子不能直接使用默认值