Enterprise
The Rockfish Workbench can be deployed in your own datacenter, you need a way to pull and run OCI or Docker containers.
It can be Kubernetes, systemd, your own solution it does not matter. Let's have a look at the architecture and the various components
App Name | Area | Short Description | Required |
---|---|---|---|
app-ui | control plane | As the name suggested a web app that can be used to interact with the system | No |
poc2 | control plane | Our HTTP API | Yes |
poc2-admin | control plane | An internal HTTP API that can be used for access management | Yes |
object store | control plane | Rockfish Workbench can manage snapshot of your data if you want | No |
SSSQL | control plane | SQL query engine to inspect datasets | No |
Redis | control plane | The control plane uses Redis as database | Yes |
cuttlefish | data plane | A worker written in Python capable of training, generate | Yes |
Redis | data plane | Our worker relay on Redis stream to exchange data in between jobs | No |
Redis
The workbench replays on Redis and its flexibility both as data pipeline (via Redis Stream) and as metadata store.
Based on capabilities and scales you can deploy two independent Redis (even in Cluster mode) to scale and operate them independently since they serve two distinct needs or you can relay on a single installation, this is why we marked the second one as not requited.
Object Store and SSQL
The Rockfish Workbench solution has a feature called dataset. Dataset are tables containing the generated value or snapshot of your data you want to train a model with.
This feature is not mandatory to use because via Data Source the workbench can do and read data where you have them. Right now we support GCP Storage, AWS S3 and Databricks.
If you need us to storage your data you have to provide an object store to use, it can be:
- AWS S3
- GCP Storage
- Azure Object Store
- MinIO
To read a dataset you can download all of that via HTTP API or you can filter
and query a subset of a dataset via SQL query, to do that we serve a container
called sssql
.
Deploy
Check our deploy page to know how we can help you releasing Rockfish Workbench.