Euler stands for Erweiterbarer, Umweltfreundlicher, Leistungsfähiger ETH-Rechner. It is an evolution of the Brutus concept. Euler also incorporates new ideas from the Academic Compute Cloud project in 2012–2013 as well as the Calculus prototype in 2013.
The Euler cluster is not destined to replace Brutus, at least not in the near future, but to complement it. Whereas Brutus is optimized for high-throughput, Euler is designed squarely for speed.
The general EULER concept was approved by the Executive Board of ETH on 1 July 2013. Due to lack of space in Zurich, the Euler cluster is physically located in CSCS's new data center in Lugano but is operated from Zurich by the HPC Group of ETH, in much the same way as the Brutus cluster today.
The first phase of Euler was delivered at the end of 2013 and installed in early 2014. The cluster was opened to all users on 30 April and inaugurated officially on 12 May. It was expanded the following month.
The second phase of Euler was delivered in December 2014. It is currently undergoing stability and performance tests and will be operational in early March 2015. This is a major expansion that will more than double the cluster's computing capacity.
The first phase of Euler contains a total of 448 compute nodes — Hewlett-Packard BL460c Gen8 —, each equipped with two 12-core Intel Xeon E5-2697v2 processors (2.7 GHz nominal, 3.0–3.5 GHz peak). All nodes are equipped with DDR3 memory clocked at 1866 MHz (64 × 256 GB; 32 × 128 GB; 352 × 64 GB) and are connected to two high speed networks (10 Gb/s Ethernet for file access; 56 Gb/s InfiniBand FDR for parallel computations).
Compared to Brutus, Euler offers:
- 3x more performance per core (28 vs 8.8 GF peak)
- 36% more performance per node (576 vs 422 GF peak)
- 30% more computing capacity overall (260 vs 200 TF)
The second phase of Euler contains 320 compute nodes of a newer generation — BL460c Gen9 —, each equipped with two 12-core Intel Xeon E5-2680v3 processors and 64 GB of DDR4 memory. These additional nodes — in production since March 2015 — increase cluster's computing capacity to approximately 570 TF.
- All nodes are connected to the cluster's Gigabit Ethernet backbone
- All nodes are connected to a high-speed InfiniBand FDR network