Systems¶
Although the broker is considered a single service, to provide its functionality it is split into applications and dependend services.
Applications¶
Web (Flask/Invenio)¶
The web service is a Flask application using the Invenio framework. It handles the REST API endpoints that receive and serve the data that the broker processes and stores. It exposes two endpoints:
/events
- Used for receiving Scholix link data in the form of events in JSON format
/relationships
- Used for serving user queries for information regarding the graph, its objects and their metadata
Worker (Celery/Invenio)¶
The worker service is responsible for asynchronously running background tasks
that are initiated either via the web application or periodically (in a UNIX
cron
-like fashion). These tasks usually involve processing events, updating
the graph state and indexing documents on Elasticsearch.
Services¶
The broker’s applications depend on some other services used mainly for sotrage and operational purposes.
PostgreSQL¶
This is the database of our choice that persistently stores the raw events and graph state we described in the Data model section.
Elasticsearch¶
This is the searchable document store that is used to store denormalized information from the graph.
RabbitMQ¶
This is the message queue used for sending tasks to Celery from the web application.
Redis¶
This is a key-value store used for various short-lived storage purposes. The web application uses its for storing rate-limiting information. The worker uses it for storing task state, task results and as a temporary cache that.