Categories
Events

Final SEA Projects Workshop

16 January 2024 – Munich, Germany A Flexible, Integrated Software Stack for HPC Systems The three EuroHPC projects DEEP-SEA, IO-SEA and RED-SEA have developed exciting European technology for use in HPC systems, targeting improvements in delivered performance and energy efficiency as well as ease of use of heterogeneous systems. The three SEA projects are now […]

Categories
Deliverables

D4.7 Optimized MPI and compute in network implementations

Edited by Simon Pickartz (ParTec) Authors Xu Huang (ParTec), Simon Pickartz (ParTec), Carsten Clauss (ParTec), Gilles Moreau (CEA), Hugo Taboada (CEA), Marc Pérache (CEA), Timo Schneider (ETHZ) Executive summary This deliverable presents improvements of the support for the BullSequana eXascale Interconnect (BXI) by the two Message-Passing Interface (MPI) implementations ParaStation MPI and Multi-Processor Computing (MPC). […]

Categories
News

RED-SEA at SC23

Impressions from the RED-SEA team at #SC23! Our last SC Conference as our project is ending in March 2024 😪

Categories
Blog

Low-latency Communication in RISC-V Clusters

Authors: Charisios Loukas, Evangelos Mageiropoulos and Nikos Chrysos, FORTH Don’t miss the demo at the end of this article! Low-latency inter-node communication is important in HPC clusters. Following the conclusion of the ExaNeSt project, which studied the adoption of low-cost, power-efficient Arm processor clusters for Exascale-class systems, as part of the RED-SEA project, we swapped […]

Categories
Events

SC23

12-17 November 2023 – Denver, CO, USA At SC23 the SEA projects will share again a booth with project ACROSS. Visit us at booth #214. Don’t miss our raffles on booth 214: several headsets to win EVERY DAY! Draws at 8:30pm on Monday, 5pm on Tuesday and Wednesday, 12pm on Thursday We are also co-organising […]

Categories
Publications

HEAR: Homomorphically Encrypted Allreduce

This paper prepared by ETH Zürich was accepted at SC23. Abstract Allreduce is one of the most commonly used collective operations. Its latency and bandwidth can be improved by offloading the calculations to the network. However, no way exists to conduct such offloading securely; in state-of-the-art solutions, the data is passed unprotected into the network. Security […]

Categories
Blog

The APEnet interconnect architecture in RED-SEA

Author: Andrea Biagioni, INFN One of the goals of RED-SEA is to tightly integrate network interfaces to different computing nodes, accelerators and workloads. We are addressing interfaces with ARM and RISC-V to prepare a number of EPI-related IPs for the new interconnect, but more generally we are designing an effective network interface for different workloads […]

Categories
Publications

Canary: Congestion-aware in-network allreduce using dynamic trees

This paper prepared by ETH Zürich was accepted in journal Future Generation of Computer Systems. Highlights Abstract The allreduce operation is an essential building block for many distributed applications, ranging from the training of deep learning models to scientific computing. In an allreduce operation, data from multiple hosts is aggregated together and then broadcasted to each host participating in the operation. Allreduce […]

Categories
Events

RED-SEA’s 4th face-to-face meeting

10-11 October 2023 – Rome, Italy On 10 and 11 October 2023 the RED-SEA team will gather for its fourth face-to-face meeting in Rome, Italy, at the invitation of consortium member INFN.

Categories
News

RED-SEA 4th F2F meeting in Rome

The fourth face to face meeting of the RED-SEA consortium is taking place in the eternal city – Rome – at the invitation of INFN. It is a great opportunity to welcome a few newcomers to the project – and for the “old hands” it is a pleasure to meet again! At 6 months of […]