SPANISH SUPERCOMPUTING NETWORK

Open

The Spanish Supercomputing Network (RES) was inaugurated in March 2007 in response to the need to increase computing capacity available to the scientific community. To address this need, the MareNostrum supercomputer was upgraded by replacing IBM JS20 blades with IBM JS21 blades, thereby doubling its computing power. The replaced blades were then used to create a distributed supercomputing infrastructure across different locations throughout Spain. Half of these blades were used to expand Magerit, the supercomputer operated by CeSViMa (UPM). The remaining blades were distributed equally to establish computing nodes at the Universities of Cantabria, Málaga, Valencia, Zaragoza, and the Instituto de Astrofísica de Canarias (IAC). As a result, the RES was originally composed of the following supercomputers:

• MareNostrum at the BSC (Barcelona Supercomputing Center).

• Magerit at CeSViMa (Technical University of Madrid, Madrid).

• Altamira at IFCA (University of Cantabria, Cantabria).

• LaPalma at the Instituto de Astrofísica de Canarias.

• Picasso at the University of Málaga.

• Tirant at the University of Valencia.

• CaesarAugusta at BIFI (University of Zaragoza).

Today, the expanded RES constitutes a distributed Unique Scientific and Technical Infrastructure (ICTS). Updated information on its member institutions and computing capabilities can be found at the following link.

Open

At the University of Valencia, the machine is named Tirant, after the protagonist of the Valencian novel Tirant lo Blanc, written by Joanot Martorell in 1490. In November 2012, MareNostrum underwent a second upgrade, becoming a 1 PFlop/s (1 Petaflop) system. Consequently, all RES nodes were upgraded, and Tirant was reconfigured with 2,048 PowerPC 970+ cores and 4 TB of distributed RAM, reaching a peak performance of 18.8 TFlops/s. Since then, each new upgrade of MareNostrum has been followed by the modernization of some of the network’s nodes. At the RES-UV node (Tirant), we currently operate Tirant version 4. With more than 13,000 cores and an Rpeak performance of 910 TFlops, it is the most powerful version of Tirant ever deployed in production. These supercomputers are accessible to any researcher. Computing time is shared between the RES, which receives 50% of the available resources in the case of Tirant, and the hosting institution, which is allocated the remaining 50%.

Open