This document describes the Workflow Invocation Ontology (wf-invocation), a simple profile of the P-plan ontology to describe how workflow steps are invoked within a workflow execution.
The latest OWL encoding of the Workflow Invocation Ontology can be found here
The Workflow Invocation Ontology is a simple OWL2 Ontology for representing the invocation of workflow specifications when they are sent to a workflow execution engine. This Ontology extends P-Plan [P-Plan], reusing some of its classes and properties.
By exposing the workflow invocation parameters, we are able to determine the commands sent to the workflow execution engine with the workflow specification, which may help to understand the provenance results in case of failure. This information is also helpful to reproduce the workflow, since we can invoke the specification in the same way as in the original workflow.
This document specifies the classes and properties of the Workflow Invocation ontology.
The Workflow Invocation Ontology extends P-Plan with two new classes (wf-invoc:Step and wf-invoc:Variable) and three data properties (wf-invoc:hasDataBinding, wf-invoc:hasCustomData and wf-invoc:hasInvocationLine). An overview of the ontology an be see in Image 1:
As shown in Figure 1, three new data properties are used to add the workflow invocation metadata to the workflow specification in P-plan. The data bindings between Variables and the actual files are linked with wf-invoc:hasDataBinding; the invocation line of the different Steps is specified through the wf-invoc:hasInvocationLine and any other metadata that the Step could require (e.g., additional tool requirements, memory restrictions, etc.) can be specified with the wf-invoc:hasCustomData.
Since the three new data properties have as domain refined p-plan classes, we have extended p-plan:Step and p-plan:Variable with wf-invoc:Step and wf-invoc:Variable respectively. These classes represent the p-plan:Steps and p-plan:Variables contextualized in a scientific workflow invocation domain.
We would like to thank Silvio Peroni for developing LODE, a framework used to build part of this web page.