Middleware for memory and data-awareness in workflows (Maestro)

Middleware for memory and data-awareness in workflows (Maestro)

Only 124 seats left
Wednesday, June 30, 2021 3:00 PM to 4:00 PM · 1 hr. (Africa/Abidjan)
HPC Workflows

Information

Contributors:
Abstract:

This poster provides an update on the work of the Maestro project. This project has the ambition of developing a middleware that enables data- and memory-aware use of High Performance Computing (HPC) and High Performance Data Analytics (HPDA) systems in the presence of complex memory and storage hierarchies. Maestro is based on the generic abstraction of Core Data Objects (CDOs) that are managed efficiently by a core middleware layer. CDOs comprise a flexible list of attributes that can be used to steer the middleware as well as the interaction between different application components and the Maestro pool that holds the set of available CDOs.

We provide an overview of the low-level core middleware and the protocol for CDO exchange between different components through a centrally managed pool. Furthermore, different middleware components are introduced together with their use by applications and workflows.

Visit the Project Website