This website uses cookies to manage authentication, navigation, and other functions. By using our website, you agree that we can place these types of cookies on your device.

View e-Privacy Directive Documents


The DEEP-EST prototype to be installed in summer 2019, will contain:


  • Cluster Module: to run codes (or parts of them) requiring high single-thread performance

  • Extreme Scale Booster: for the highly- scalable parts of the applications

  • Data Analytics Module: supporting HPDA requirements

The three mentioned compute modules will be connected with each other through a “Network Federation” to efficiently bridge between the (potentially different) network technologies of the various modules. Attached to the “Network Federation,” two innovative memory technologies will be included:

  • Network Attached Memory: providing  a large-size memory pool globally accessible to all nodes

  • Global Collective Engine: a processing element at the network to accelerate MPI collective operations

In addition to the three above mentioned compute modules, a service module will provide the prototype with the required scalable storage.


One important aspect to be considered in the design and construction of the DEEP-EST prototype is energy efficiency. It will influence the choice of the specific components and how they are integrated and cooled. An advanced monitoring infrastructure will be included to precisely quantify the power consumption of the most important components of the machine, and modelling tools will be applied to predict the consumption of a large scale system built under the same principles.


The basis for the MSA: The Cluster-Booster Architecture

The development of the underlying architectural concept started in DEEP and DEEP-ER with the
Cluster-Booster approach. This architecture integrates two different HPC systems:

  • Standard Cluster: general purpose multi-core processors  (Intel® Xeon®) with high single-thread performance.
  • Extreme Scale Booster: many-core processors (Intel® Xeon® Phi™) with leading aggregated
    performance on vectorizable  codes and high scalability  and energy efficiency.