Workflows
Workflows
In PerMedCoE, building blocks are used to package core simulation tools and other ancillary processes, such as data analytics. They help unify the interfaces to these elements so that it becomes easier to compose them into workflows and to deploy them in HPC resources. Even though Building blocks can be used stand-alone, they are intended to be composed into workflows. In particular, the PerMedCoE is implementing workflows to solve the different use cases of the project.
‘PermedCoE’ Python Package
A python package, called ‘permedcoe’, has been developed to develop and run building blocks and workflows.
Our goals
PerMedCoE developments will be based on containers enabling reproducible operations in heterogeneous HPC infrastructures, and their inclusion into efficient building blocks and workflows, reproducibly encapsulating the full chain of operations required in this area. PerMedCoE will carry out implementations in the main workflow platforms and will build systems of workflows optimised/adapted to the necessary complexity level to satisfy the requirements of different types of users, represented by use cases. The PerMedCoE methodology will follow the FAIR (Findable, Accessible, Interoperable and Reusable) principles and the recommendations for the development of software containers and interoperable workflows elaborated by ELIXIR (using the ELIXIR repository, co-led by BSC) and other bodies (GA4GH).
The workflows will involve innovative AI/ML methods to:
- Facilitate the selection/estimation of parameters for complex simulation environments;
- Monitor and redirect computational efforts in large simulations; and
- In conjunction with data analytics techniques, develop flexible analytical environments to deal with the resulting large-scale simulation results. PerMedCoE will be interested in leading AI/ML developments and adapting complementary methods, as soon as their applicability is consolidated.