Here is a brief architecture of Jubakit:
There are 6 components that consists of Jubakit:
- Loader fetches data from various data sources (e.g., CSV file, RDBMS, MQ, Twitter stream, etc.) in key-value format.
- Schema defines the data type (string feature, numeric feature, ground truth (label), etc.) for each keys of data loaded by Loader.
- Dataset transforms records loaded from Loader into Jubatus Datum using Schema. Dataset is an abstract representation of a sequence of data.
- Service makes update/analyze RPC calls to Jubatus servers for each record in Dataset and returns the result.
- Config defines parameters of Service.
- Shell provides an interactive command line interface to communicate with Jubatus servers.