The DEEP Project An alternative approach to heterogeneous cluster-computing in the many-core era

Norbert Eicker, Thomas Lippert, Thomas Moschny, Estela Suarez
2015 Concurrency and Computation  
Homogeneous cluster architectures, which used to dominate high-performance computing (HPC), are challenged today by heterogeneous approaches utilizing accelerator or co-processor devices. The DEEP (Dynamical Exascale Entry Platform) project is implementing a novel architecture for High-Performance Computing, in which a standard HPC Cluster is directly connected to a so-called "Booster": a cluster of many-core processors. By these means heterogeneity is organized differently as in today's
more » ... d approach, where accelerators are added to each node of the Cluster. In order to adapt application codes to this Cluster-Booster architecture as seamless as possible, DEEP has developed a complete programming environment. It integrates the offloading functionality given by the MPI standard with an abstraction layer based on the task-based OmpSs programming paradigm. This paper presents the DEEP project with an emphasis on the DEEP programming environment. Early examples are dedicated devices from ClearSpeed, ranging over processors originally developed for gaming like the Cell Broadband Engine and the aforementioned GPGPUs, towards newer developments like Intel's Xeon Phi. ¶ We will distinguish between accelerated clusters, i.e. classical clusters comprising of nodes equipped with accelerators, and clusters of accelerators, i.e. clusters connected to each other in an own entity such as in the Booster concept introduced by DEEP. In principle the Cluster-Booster approach doesn't put any constraints on the Booster network topology. Since the Booster is claimed to be scalable, its fabric must have a scalable topology like torus, butterfly, etc.
doi:10.1002/cpe.3562 fatcat:3xeoqn7xtzekzkmk36dhpblci4