Computing Fat Nodes

Fat node usually uses a ccNUMA architecture. All memory is visible to and accessible from any CPU and cache coherency is handled in hardware. Fat node features outstanding single-node performance, extremely high memory capacity, ultra-high reliability to satisfy the needs of memory-hungry or mission-critical applications.

99.999% Reliability

Full Redundancy

Typical Applications

Life Science

  • Sequence assembly
  • Multi-threaded, memory-demanding and long-running

Numerical Weather Forecasting

  • Cron jobs including pre- and post-processing
  • Mission-critical and high-throughput

CAE

  • Implicit FEM analysis
  • Memory and disk I/O intensive
  • Limited scalability

CAE

  • Electromagnetic simulation
  • Compute-intensive
  • Memory-hungry

Technical Specifications

I840-G25

Rack mount 4U
4x Intel Xeon processor E7-4800 v2/v3 family
96x DDR3 DIMM slots, up to 6TB memory
24x Hot-swap 2.5“ SAS/SATA HDD or SSD
11x PCI-E 3.0 slots, up to 4 GPUs or Intel Xeon Phis
Optinal: FDR 56Gbps or EDR 100Gbps InfiniBand
4x PSUs, N+N redundancy
I980-G10
Rack mount 8U
8x Intel Xeon processor E7-8800 v2/v3 family
128x DDR3 DIMM slots, up to 8TB memory
16x Hot-swap 2.5“ SAS/SATA HDD or SSD with RAID 0/1/5/6
16x PCI-E 3.0 slots, up to 4 GPUs or Intel Xeon Phis
Optinal: FDR 56Gbps or EDR 100Gbps InfiniBand
4x PSUs, N+1/N+N redundancy