Cerise

• Lets you make running workflows on HPC resources easy
• You provides users with a CWL API, they combine tools into a workflow
• Installs software, accepts workflows, stages files, executes
• REST/CWL/Docker/Python client

Cerise is a generic service for running CWL workflows on compute resources (i.e. clusters, supercomputers, and simply remote machines). It tries to offer a consistent environment for workflows, so that a workflow sent to resource A will work unchanged on resource B as well.

To achieve this, and to offer a bit of safety and perhaps security, Cerise does not allow running arbitrary CWL command line tools. Instead, it expects the user to submit a workflow document that refers to predefined steps built into the service.

Defining these steps, and adding them to the service, is called specialising the service. A specialisation of Cerise is always specific to a project (which determines which steps are available and what inputs and outputs they have), and to a compute resource (which determines how they are implemented). Thus, two workflows sent to two different specialisations to the same project, but to different compute resources, should give the same result (assuming the calculation is deterministic!).