The Glenn cluster is built on AMD Opteron 6220 (code-named "Interlagos") CPU's. The system consists of:
In total 379 compute nodes (total of 6080 cores) and 18.1 TB of RAM. More specific:
224 nodes with 16 cores and 32 GB of RAM
135 nodes with 16 cores and 64 GB of RAM
13 nodes with 16 cores and 128 GB of RAM
1 node with 32 cores and 512 GB of RAM
4 nodes with 16 cores, 32 GB of RAM and 1 NVidia Fermi M2050 GPU
2 nodes with 16 cores, 32 GB of RAM and 1 PCoIP adapter for remote graphics
There are also 3 system servers used for accessing and managing the cluster.
There's a Gigabit Ethernet network used for logins and file system access, a dedicated management network and an Infiniband high-speed/low-latency network for parallell computations. The nodes are equipped with Mellanox ConnectX-2 QDR Infiniband 40Gbps HCA's.
The server and compute node hardware is built by Supermicro and delivered by South Pole.
Abisko
HPC2N
6 030
—
4 500
x 1000 core-h/month
The cluster has 15744 cores with a peak performance of over 150 Tflops/s. For high parallel performance, the system is equipped with a high bandwidth, low latency QDR InfiniBand interconnect, with full bisectional bandwidth. All nodes have at least 2 GB/core and some nodes have over 8 GB/core. For more information about the system and available software see the HPC2N web-pages.
Akka
HPC2N
1 030
—
800
x 1000 core-h/month
The cluster contains 672 nodes each in a dual quad-core configuration, for a total of 5376 cores. The CPU:s are the Intel L5420 low-power CPU:s running at 2.5 GHz, giving each node a total peak performance of 80 Gflops/s. Each node is equipped with 16 GB RAM for a total of more than 10 Terabytes of RAM. For high parallel performance, the system is equipped with a high bandwidth, low latency InfiniBand interconnect. Scratch storage is provided by a 100 TB disk-subsystem running GPFS. For more information about the system and available software see the HPC2N web-pages.
Akka will be closed down at the end of 2012. In this round we will only award 6 months allocations.
Alarik
LUNARC
200
—
450
x 1000 core-h/month
Alarik was cluster consisting of 208 nodes. Each nodes had 2 eightcore CPU:s with 32-64 GB memory and 250 GB local scratch. The interconnect was Infiniband with 40 Gb/s. The system was connected to a gigabit Ethernet.
Alarik has been decomissioned.
Triolith
NSC
13 775
—
5 000
x 1000 core-h/month
Triolith (triolith.nsc.liu.se) was a capability cluster with a total of 24320 cores and a peak performance of 428 Tflops/s. However, Triolith was shrunk by 576 nodes on April 3rd, 2017 as a result of a delay in funding a replacement system and now has a peak performance of 260 Teraflop/sec and 16,368 compute cores. It is equipped with a fast interconnect for high performance for parallel applications. The operating system is CentOS 6.x x86_64. Each of the 1520 (now 944) HP SL230s compute servers is equipped with two Intel E5-2660 (2.2 GHz Sandybridge) processors with 8 cores each (i.e. 16 cores per compute server). 56 of the compute servers have 128 GiB memory each and the remaining 888 have 32 GiB each. The fast interconnect is Infiniband from Mellanox (FDR IB, 56 Gb/s) in a 2:1 blocking configuration.
Triolith have been replaced with a new system, Tetralith, that was made available to users on August 23, 2018. NSC currently plan to keep Triolith in operation and available to users until September 21st, 2018. After that, Triolith will be permanently shut down and decommissioned.
Lindgren
PDC
12 840
—
7 900
x 1000 core-h/month
Kalkyl
UPPMAX
300
—
180
x 1000 core-h/month
Kalkyl will be retired 2013-12-31.
Kalkyl will be retired 2013-12-31. Time will be allocated for a maximum of 6 months.
The Linux-based capacity cluster consists of 348 compute servers, each server with two Quad-core Intel® Xeon® E5520 (Nehalem 2.26 Ghz, 8MB cache) processors, 24 GB memory and 250 GB local disc. We have increased the memory of 16 of these servers to 48 GB. 16 more of the servers have their memory increased to 72 GB and their local disk space increased to two TB. An InfiniBand network interconnects all compute servers, with a bandwidth of 20 Gb/s.
Tintin
UPPMAX
650
—
360
x 1000 core-h/month
160 compute servers, each with two AMD Opteron 6220 (3.0 GHz 8 core Bulldozer) CPUs. 144 of the nodes have 64 GB memory and 16 have 128 GB memory. All nodes have 1 TB local disc. All nodes are interconnected with a 2:1 oversubscribed QDR Infiniband fabric. There is also 4 nodes with Tesla GPU graphics card. SNICs share of the Tintin system is 85.9%, i.e. about 137 compute servers.
Click above to show more information about the resource.
