2015 has seen several increases of the computing capacity of the UL HPC platform with the following additions to the Gaia cluster which more than doubled its computing power:
- Delta D88x-M8-BI: gaia-80
- 8 * Intel Xeon E7-8880v2 @ 2.5 GHz (120 cores: 2.4 TFlops) with 3TB of RAM and 12.6TB of local storage;
- SGI UV 2000: gaia-81
- 16 * Intel Xeon E5-4650v2 @ 2.4 GHz (160 cores: 3 TFlops) with 4TB of RAM; The large amount of memory available on these systems enable in-memory processing of huge datasets, making possible amazing analyses and simulations.
Three ultra-dense enclosures with state-of-the-art general purpose computing systems:
- Dell FC430: gaia-[155..178]
- 24 computing nodes, each with 2 * Intel Xeon E5-2680v3 @ 2.5 GHz and 128GB of RAM (576 cores: 23.04 TFlops)
Finally, two dense, energy efficient enclosures with low power high performance CPUs based on the Haswell architecture featuring the Crystalwell package (large 128MB L4 cache), ideal for a variety of BigData analytics workloads that are loosely coupled:
- HP Moonshot: moonshot1-[1..45], moonshot2-[1..45]
- 90 computing nodes, each with an Intel Xeon E3-1284Lv3 @ 1.8 GHz and 32GB of RAM (360 cores: 10.39 TFlops and 2.88 TB of RAM in total).
Due to their unique properties, user jobs are not scheduled by default on the Delta, SGI or HP nodes. Running tasks on these machines require the use of specific settings, see the UL HPC OAR documentation for bigmem/bigsmp and respectively moonshot systems.