Promoting a Standards-Based, Abstraction-Driven Approach to Distributed Computing
RADICAL Cybertools is an abstractions-based suite of well-defined capabilities that are architected for scalable, interoperable and sustainable approaches to support science on a range of high-performance and distributed computing systems.
It currently consists of two components: RADICAL-Pilot: a scalable and flexible Pilot-Job system that provides flexible application-level resource management capabilities, and RADICAL-SAGA: a lightweight interface that provides a standards-based interoperability across a range of computing systems. We will add more components such as, advanced "overlay" based workload management systems (RADICAL-OWMS) in the near future.
RADICAL Cybertools builds upon important theoretical advances, production software development best practices and well defined usage and programming models. RADICAL Cybertools is also a vehicle for active cyberinfrastructure research, which enable its users to seamlessly reach the next level of scale and functionality.
RADICAL-SAGA is the infrastructure access layer of the RADICAL Cybertools stack. It provides a homogeneous programming interface to the majority of production HPC queuing systems, Grid- and Cloud-services. RADICAL-SAGA supports XSDE and OSG resources as well as academic and commercial cloud computing platforms.
RADICAL-Pilot (also known as BigJob) is the infrastructure abstraction layer of the RADICAL Cybertools stack. It provides a programming library to create ad-hoc resource overlay and multilevel scheduling environments based on Pilot-Jobs. RADICAL-Pilot can marshal and O(10k) distributed cores and schedule O(10k) tasks in a single application context across multiple distributed distributed HPC resources.
RADICAL-OWMS is the infrastructure federation layer of the RADICAL Cybertools stack. It provides programming libraries to create ad-hoc Workload Management Systems (WMS) and Overlay Management Systems (OMS) to execute application workloads, like bag of tasks, ensembles, and workflows on multiple distributed computing infrastructures. The RADICAL-OWMS libraries provide support for tailoring execution strategies to the workload requirements, resource selection within and across DCI, creation and handling of compute and data units, creation and management of Pilot-Jobs, efficient, "intelligent" data handling.
The EnsembleMD Toolkit (EnMDTK) project is a loose collection of RADICAL-Pilot based tools, programming libraries and APIs that simplify and abstract common tasks, patters and tools for molecular dynamics workflows. All of these tools encapsulate at least one of the commonly used MD engines, GROMACS, AMBER and NAMD.