Workflow Execution Sample Clauses
Workflow Execution. 1 The system should allow re-processing inner steps of a workflow being executed, at the request of a user. For instance, in case the output of a certain task/execution step is not satisfactory, user should be able to modify some of its execution parameters and/or reprocess it on demand.
Workflow Execution. Workflow execution involves four main components: The workflow executor as core component, which orchestrates and uses the other components and is based on the Apache Camel framework [Fou04]. The RabbitMQ message broker, which is used as communication layer to loosely couple extractors and the MICO platform [PS04]. The micoRabbitComp, which is a Camel endpoint component that connects the Camel framework with the MICO platform, and triggers the extractors via RabbitMQ. The auxiliary componentm which is also a MICO-specific extension to Camel that allows extrac- tion of information from Marmotta to support dynamic routing based on information provided by extractors of the MICO platform. involves several extractors, the auxiliary component, and a dynamic router to complete the task. The connections between components and their execution order is depicted in figure 15. The first extractor audio demux extracts the audio stream from an injected video file and stores it within platform storage for further processing. The next two extractors are diarization and language detection, which are analyzing the audio content in parallel. Finally, the diarization extractor annotates information about the speaker and tries to detect start and end points of sentences, and language de- tection is used to identify the spoken language, storing it in the knowledge base. Afterwards, a Camel component loads that data from storage, and puts it to the Camel message, before a Camel component for dynamic routing checks the language information and triggers a Kaldi extractor that is configured with the language model according to the detected language.
Workflow Execution. Workflow execution takes an actionable workflow to be executed (from the WF-Exec space in Figure 3) and translates the elements related to each allocated resource into resource invocation data. Depending on the resource to be invoked, the workflow execution service either transmits the data to the resource, or if the resource needs to be started, starts the resource with the appropriate configuration. Each workflow execution service provider manages a group of operational resources, ranging from all operational resources (fully centralized model, see Figure 13) to one resource (fully distributed model, see Figure 14).