DEEP / DEEP/-ER / DEEP-EST

The DEEP, DEEP-ER and DEEP-EST projects are an innovative European response to the Exascale challenge. The consortium develops a novel, Exascale-enabling supercomputing architecture with a matching software stack and a set of optimized grand-challenge simulation applications.

DEEP takes the concept of compute acceleration to a new level: instead of adding accelerator cards to Cluster nodes, an accelerator Cluster, called Booster, will complement a conventional HPC system and increase its compute performance. Together with a software stack focused on meeting Exascale requirements - comprising adapted programming models, libraries and performance tools - the DEEP architecture will enable unprecedented scalability. 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. An extrapolation to millions of cores would take the DEEP concept to an Exascale level.

DEEP-ER advances the Cluster-Booster architecture developed in DEEP from a hardware point of view in terms of processor technology, network interconnect, and storage. On the software side, the project focuses on two central research topics: highly scalable parallel I/O and resiliency.

The Modular Supercomputer Architecture (MSA) developed in the DEEP-EST research project is a blueprint for large-scale heterogeneous systems supporting the different requirements of HPC, Big Data Analytics and Machine Learning with highest efficiency and scalability. The MSA is based on the Cluster-Booster architecture developed in the previous DEEP and DEEP-ER projects.

In DEEP, LRZ contributed with its expertise in energy efficiency to hardware design, system software, and operations. Within the DEEP-ER project, LRZ involved with a demonstration application code from the field of earthquake simulations. LRZ also manages the project’s dissemination activities ensuring the projects’ visibility to the HPC community, the European Commission and the general public through a variety of communication channels.

Highlights

With respect to DEEP-ER, LRZ focuses on the I/O developments to their earthquake source dynamics code and has made substantial progress in 2016.

As far as DEEP is concerned, the full prototype of 500 TFLOP/s peak performance has been up and running at Jülich Supercomputing Centre since end of 2015. In parallel the Energy Efficiency Evaluator (EEE), which is a smaller replica of the system in Jülich, was installed at Leibniz Supercomputing Centre to implement and advance the fine-grained DEEP energy monitoring system.

Events and Conferences

SC16, November 13 – 18 2016, Salt Lake City (UT) / United States

ISC’16, June 19 – 23 2016, Frankfurt/Germany

European HPC Summit Week, May 9 – 12 2016, Prague / Czech Republic

CeBIT, March 14 – 26 2016, Hannover / Germany

Fast Facts