MuPIF DB & scheduler¶
MuPIF DB is the data management system (DMS) of the platform. It keeps the Digital Twin and records how every piece of simulated data was produced.
-
Workflow registry
Workflows are uploaded with their metadata and versioned. Their inputs and outputs are described in the metadata, so any registered workflow can be configured and executed from the web interface.
-
Execution scheduler
The scheduler picks up scheduled executions, allocates the models through the job managers, runs the workflows and stores their results and logs. It also collects statistics on the platform load.
-
Digital Twin database
The Entity Data Model is generated from your JSON schema and stored in MongoDB. Workflow inputs are read from the twin and outputs create new entity states.
-
REST API and web interface
Everything is accessible through a REST API, with a "safe" subset for public access, and through a web interface for browsing workflows, setting up executions and inspecting their results.
-
Traceability and reproducibility
Each execution stores its inputs, outputs, the workflow version and the logs. Any result can be traced back and reproduced.
-
Monitoring
The MuPIF monitor shows the status of all platform services and the history of recent executions.
Life of an execution¶
- A user selects a registered workflow and an entity of the Digital Twin, and creates an execution.
- Inputs are filled in from the twin through EDM paths, or set manually.
- The execution is scheduled. The scheduler starts the workflow, which allocates its models on the distributed platform.
- When the workflow finishes, its outputs are stored and written to new entity states of the twin, together with the execution record and logs.
Sources and documentation: github.com/mupif/mupifDB.