Supercomputer (HPC)

Our HPC service is focused on solving situations of great impact that occur in science and industry. The infrastructure that we offer together with our qualified personnel results in a better use of architecture and less processing time.

An additional value to our service is the timely advice of scientists and experts on HPC issues, which will help us find solutions and improvements to problems. The following options are available:


SIMULATION
Focused on showing results based on program outputs to represent data graphically.

ANALYSIS
Analysis of large blocks of information with countless variables to find their behavior.

Bigdata
Analysis of large blocks of information with countless variables to find their behavior.

Our supercomputer
Thubat Kaal 2.0

257.24 Bull Equipment Theoretical Teraflops

Computation capacity of 257,243 TERAFLOPS in double precision (182.04 Intel Skylake Xeon x86 64 Bits, 75.2 NVIDIA GPU P100). Composed of 86 calculation nodes of which 82 nodes are INTEL XEON X86 BITS SKYLAKE (Purley Platform) technology and 4 nodes contain 4 NVIDIA P100 c/u boards. It has a High Availability Luster file system with 1.8 PB available. GREEN TECHNOLOGY and environmentally friendly. GREEN TECHNOLOGY and environmentally friendly.

Hardware
4 RACK 42 U with compatible PDUs and power cables for equipment. 86 expandable calculation nodes with the following characteristics:

64 BULL R440-E5 compute nodes
  • 2 Intel Xeon Skylake processors 6130 16c, 2.1 - 3.7 ghz (32 CORES per node)
  • 192 GB RAM (6 GB per CORE)
18 BULL R440-E5 compute FAT nodes
  • 2 Intel Xeon Skylake processors 6130 16c, 2.1 - 3.7 ghz (32 CORES per node)
  • 384 GB RAM (12 GB per CORE)
4 GPU BULLX R421-E4K nodes
  • 2 INTEL XEON E5-2680 processors 14c, 2.4 GHZ (28 CORES per node)
  • 128 GB RAM (4.57 GB per core)
  • 4 NVIDIA P100 cards per node.
2 administration nodes for services configured in high availability cluster mode (HA Active-Active)

2 access nodes for user login services and home space configured in active high availability cluster mode (HA Active- Active) for login in Active-Passive mode (HA-Active-Pasive) for home space. 200 TB /home directory space.

High performance Infiniban EDR 100GBPS network configured in an All to All topology.
This makes us the fastest infiniband cluster in Mexico today.
10 Gbps Ethernet Management Network

Lustra file system for Scratch 1.8 PB or 1843 Tb approx.

Software
Operating system: RedHat 7.3 licensed throughout the cluster.
Scheduling system: Slurm.