Installation Manual - Networks
Introduction
DQ0 uses multiple networks to ensure a safe environment. At least two networks should be set up in DQ0:
The private network, which for example corresponds to the customer's internal network and is not accessible from "outside". This network contains the confidential data that can be made available for machine learning, as well as the DQ0 instance itself, which consists of the platform and the runtime environment.
The public network, on the other hand, is freely accessible and can be used to allow users outside the private network to access the DQ0 instance.
Public Network
| Application | component | Protocol | Direction | Description |
|---|---|---|---|---|
| cli | dashboard | TCP/8080 | Inbound | cli frontend running on port 8080 |
| cli | dashboard | TCP/9000 | Outbound | cli frontend uses internal api via Rest API / WebSockets |
| cli | api | TCP/9000 | Inbound | cli api running on port 9000 |
| cli | internal | TCP/8000 | Outbound | cli communicates with proxy (full-duplex communication using WebSocket) |
| proxy | internal | TCP/8000 | Inbound | WebSocket proxy server running on port 8000 |
Private Network
| Application | component | Protocol | Direction | Description |
|---|---|---|---|---|
| platform | internal | TCP/8000 | Outbound | platform communicates with proxy (full-duplex communication using WebSocket) |
| platform | internal | TCP/5000 | Outbound | platform communicates with tracking_server |
| service | internal | TCP/8000 | Outbound | plugins (internal computation entities) are pushing results to proxy |
| service | internal | TCP/5000 | Outbound | plugins (internal computation entities) are pushing results to tracking_server |
| tracking_server | internal | TCP/5000 | Inbound | tracking server (mlflow) |