It seems likely that for self-hosting Phoenix, the recommended specs can vary based on workload. Internally, Phoenix uses 2GB RAM and a single CPU for hosted instances, but this might not be sufficient for high workloads like 30k traces a day. You might need to scale up the number of instances or resources as you encounter performance issues. Additionally, Phoenix is stateless, so scaling out by deploying multiple containers backed by a single Postgres database is possible, which might help with performance issues (<https://arize-ai.slack.com/archives/C04R3GXC8HK/p1727401604.563959|source>).
If you're experiencing slowness, it might also be related to the database configuration or the way data is being queried. For instance, performance issues have been reported when querying data beyond a certain time interval, and optimizations like lazy loading of spans and reducing page size have been suggested (<https://github.com/Arize-ai/phoenix/issues/6201|source>). Would you like more details on optimizing your setup or specific configurations?
Sources:
- <https://arize-ai.slack.com/archives/C04R3GXC8HK/p1727401604.563959|https://arize-ai.slack.com/archives/C04R3GXC8HK/p1727401604.563959>
- <https://github.com/Arize-ai/phoenix/issues/6201|https://github.com/Arize-ai/phoenix/issues/6201>
