dApollo: Dynamic Workflow Scheduling

dApollo is a dynamic workflow scheduling algorithm for distributed runtime systems across the Edge-Cloud continuum. dApollo is integrated in the distributed runtime system Apollo.

Requirements

• Workflow in AFCL format
• Available resources for each task (typemappings)
• Scheduler configuration input, including resource bandwidth, acquisition delay of each resource and cost limit for the overall schedule

Features

• Open-source
• Targets the Edge-Cloud continuum
• Optimizes for runtime while adhering to a given cost limit
• Dynamically maps tasks to resources and orchestrates groups of workflow tasks to effectively use distributed runtime system instances
• Reacts to dynamically updated task runtimes to take resource performance fluctuations into account

Getting started

Please refer to the documentation of the Apollo Demo to get started with Apollo. The dApollo scheduling Algorithm currently supports simulated executions. The algorithm was evaluated using four well known scientific workflows: Montage, CasaWind, Genome1000 and PSLoad.