DEEP is an innovative European response to the Exascale challenge. The consortium developed a novel, Exascale-enabling supercomputing architecture with a matching software stack and a set of optimized grand-challenge simulation applications. Being its predecessor project, it lays the basis for DEEP-ER.
DEEP takes the concept of compute acceleration to a new level: instead of adding accelerator cards to Cluster nodes, an accelerator Cluster, called Booster, complements a conventional HPC system and increases its compute performance. The Cluster-level heterogeneity of DEEP attenuates the consequences of Amdahl’s Law allowing users to run applications with kernels of high scalability alongside kernels of low scalability concurrently on different sides of the system, avoiding at the same time over and under subscription. Find details on the Cluster-Booster-architecture here.
DEEP-ER builds on this core architectural concept and develops it further in terms of processor technology, network interconnect and storage. An in-depth overview on the DEEP-ER hardware extensions is presented here.
Surmounting software challenges
Programming heterogeneous systems is often considered challenging and cumbersome. Therefore, the DEEP team came up with a software stack focused on meeting Exascale requirements: Comprised of adapted programming models, libraries and performance tools the DEEP software stack enables unprecedented scalability. With its programming model, based on OmpSs, the DEEP system provides a dedicated development and runtime environment, which makes porting applications to DEEP a snap. Find details on OmpSs and Global ParaStation MPI as used in DEEP here.
Driver to Exascale: Applications
From the very beginning six real-world HPC applications have been involved in the DEEP project. Their requirements both for the hardware and system software/API aspects drive the co-design process: The selected applications have been continuously optimized for the Cluster-Booster architecture and for the CPU technologies used (Intel Xeon and Intel XeonPhi). The end result is a system that fits the needs of the applications, and application showcases that demonstrate the efficiency and performance of the DEEP systems. You find an overview on the DEEP applications here.
In DEEP-ER we continue this successful co-design approach focusing on the I/O and resiliency requirements of these seven HPC applications.