Content | Navigation |

Hardware

CCAST provides multiple general access computing resources:

Thunder Cluster

The Thunder cluster consists of 63 compute nodes and 2 login nodes. The total aggregated theroretical peak performance is ~40TFLOPS. All these nodes are interconnected with FDR Infiniband at a 56Gbit/s transfer rate. The system details are shown below

  • Compute node: 
    • 58x Dual socket Intel Xeon 2670v2 "Ivy Bridge" (10 core per socket) 2.5GHz with 64GB DDR3 RAM at 1866MHz and
      • 14x of these nodes are equipped with Intel Phi (aka mic) accelarator cards with 60 x86 cores at 1.047GHz, 7.5GB of RAM.
    • 2x Large memory nodes 1TB RAM at 1600MHz, 4 sockets with 8 cores (Intel Xeon 4640 Sandy Bridge processrs) on each socket, 3.3 TB of local SSD scratch.
    • 2x Development nodes with Intel Ivy Bridge processors with Intel Phi cards and one Sandy Bridge node.
    • 1x Mellanox ConnectX3 FDR IB on all 63 compute nodes.
  • Login Node: 2x Intel Ivy Bridge ES-2670 v2 (2 socket, 10 core per socket) 2.5GHz, 64GB DDR3 1866MHz memory, 2x 1TB SATA hard disks, 1 Mellanox ConnectX3 FDR IB, 1 dual 10GbE port.
  • Number of Compute node: 63
  • Number of Processor core: 1280
  • Number of Login node: 2
  • Storage System: 2 tier IBM GPFS filesystem with a policy driven HSM.
    • Tier 1 with 120x 10K SAS drives at about 4GB/s.
    • Tier 2 with 80x 7.2K SATA drives at about 2.6GB/s.
    • Tape storage: TS3584 L53 with 8x LTO6 Tape drive, 274x tape slot and TS3584 S54 with 1340x tape slots. Included are 258x LTO6 tape cartridge for 645TB, expandable up to 4PB.
  • Interconnection: FDR Infiniband at 56Gbit/s
  • Network Switch: Mellanox FDR IB switch

Cluster 2

The cluster2 consists of 32x compute nodes and 1x login node. The total aggregated theroretical peak performance is ~ 2.8TFLOPS. The aggregated memory and aggregated local storage are ~ 1.0TB and ~ 5.3TB, respectively. Each node is interconnected through Gigabit Ethernet switches. The system details are shown below:

  • Compute node:  dual-socket Intel E5430 2.66GHz with 32GB ECC DDR2 667MHz (2GB DIMMs), 160GB 7.2K RPM SATA HDD, dual Gigabit Ethernet ports.
  • Login node: dual-socket Intel E5405 2.00GHz with 4GB ECC DDR2 667MHz (2GB DIMMs), 2x 146GB 15K RPM SAS HDD, 3x Gigabit Ethernet ports.
  • Processor: quad-core Intel E5430 @ 2.66GHz (for Compute node), quad-core Intel E5405 @ 2.00GHz (for Login node).
  • Number of Compute node: 32
  • Number of Processor core: 256
  • Number of Login node: 1
  • Global scratch storage system: IBM GPFS
  • Global scratch storage maximum bandwidth: 1.25GB/s
  • Interconnection: Gigabit Ethernet
  • Network switch:  Cisco 3750G, Force10 S60

Cluster 3

The cluster3 consists of 128x compute nodes and 1x login node. The total aggregated theroretical peak performance is ~ 10.9TFLOPS. The aggregated memory and aggregated local storage are ~ 6.1TB and ~ 20.5TB, respectively. Each node is interconnected via both ultra-low latency high-speed Myri-10G and Gigabit Ethernet. The system details are shown below:

  • Compute node:  dual-socket Intel X5550 2.67GHz with 48GB ECC DDR3 1333MHz (8GB DIMMs), 160GB 7.2K RPM SATA HDD, 1x Myri-10G port, dual Gigabit Ethernet ports.
  • Login node: dual-socket Intel X5550 2.67GHz with 48GB ECC DDR3 1333MHz (8GB DIMMs), 160GB 7.2K RPM SATA HDD, 1x Myri-10G port, dual Gigabit Ethernet ports.
  • Processor: quad-core Intel X5550 @ 2.67GHz.
  • Number of Compute node: 128
  • Number of Processor core: 1024
  • Number of Login node: 1
  • Global scratch storage system: IBM GPFS
  • Global scratch storage maximum bandwidth: 5.0GB/s
  • Interconnection: Myri-10G, Gigabit Ethernet
  • Network switch:  Myri-10G, Cisco 3750G.

Server 1

The theoretical peak performance is ~4.2TFLOPS.  The system specifications are show below:

  • Processor: quad-core Intel X5560 @ 2.80GHz.
  • Memory: 96GB ECC DDR3 1333MHz (8GB DIMMs)
  • Local storage: 2x 146G 15K RPM SAS HDD
  • GPU: Nvidia Tesla S1070
  • Number of GPU: 4
  • Global scratch storage system: IBM GPFS
  • Global scratch storage maximum bandwidth: 1.25GB/s
  • Interconnection: Gigabit Ethernet
  • Network switch:  Cisco 3750G, Force10 S60

Student Focused. Land Grant. Research University.

Follow NDSU
  • Facebook
  • Twitter
  • RSS
  • Google Maps

CCAST Support
Phone: 701.231.5184
Physical/delivery address:  1805 NDSU Research Park Drive/Fargo, ND 58102
Mailing address:  P.O. Box 6050—Dept. 4100/Fargo, ND 58108-6050
Page manager: CCAST

Last Updated: Friday, July 28, 2017 8:31:08 AM
Privacy Statement