The Nvidia CUDA development kit has been updated to version 4.1.
DAS-4/ASTRON has succesfully correlated three hours of data coming from 288 LOFAR antennas, possibly setting a world record.
The GPU page documents how the OpenCL implementations on DAS-4 can be used.
DAS-4/VU achieved the 16th place on the Graph500 list. A very good result, as all higher-ranked systems are significantly bigger.
DAS-4 supports the OpenNebula package for Cloud computing.
DAS-4 includes roughly 200 dual-quad-core compute nodes
(primarily SuperMicro 2U-twins with Intel E5620 CPUs),
spread out over six clusters, located at five sites.
The system has been built by
ClusterVision.
Like its predecessor, DAS-3,
DAS-4 is rather heterogeneous in design:
| Cluster | Nodes | Type | Speed | Memory | Storage | Node HDDs | Network | Accelerators |
|---|---|---|---|---|---|---|---|---|
| VU | 74 | dual quad-core | 2.4 GHz | 24 GB | 30 TB | 2*1TB | IB and GbE | 16*GTX480 + 2*C2050 |
| LU | 16 | dual quad-core | 2.4 GHz | 48 GB | 50 TB | 5*2TB + .5TB SSD | IB and GbE | future upgrade (FPGAs) |
| UvA | 16 | dual quad-core | 2.4 GHz | 24 GB | 30 TB | 1TB | IB and GbE | future upgrade |
| TUD | 32 | dual quad-core | 2.4 GHz | 24 GB | 18 TB | 2*1TB | IB and GbE | 8*GTX480 |
| UvA-MN | 36 | dual quad-core | 2.4 GHz | 24 GB | 30 TB | 2*1TB | IB and GbE | 8*GTX480 + 7*C2050 + 2 x GTX480 |
| ASTRON | 24 | dual quad-core | 2.4 GHz | 24 GB | 24 TB | 1*1TB | IB and GbE | 8*GTX580 + 1*C2050 + 1*HD6970 |
Besides using the ubiquitous Ethernet (1 Gbit/s at the compute nodes, 10 Gbit/s on the head nodes), DAS-4 also employs the high speed InfiniBand interconnect technology. Quad Data Rate (QDR) InfiniBand (IB) is used as an internal high-speed interconnect, but also to connect with the other DAS-4 clusters by means of routers that communicate over dedicated 10 Gbit/s Ethernet lightpaths. These lightpaths are implemented via a fully optical (DWDM) backbone in the Netherlands, SURFnet6. See also: DAS-4 Connectivity.
The operating system the DAS-4 runs is CentOS Linux. Cluster management is done using Bright cluster management suite. In addition, software from many sources is available to support research on DAS-4: the Grid Engine resource management system, various MPI implementations (e.g., OpenMPI and MPICH), optimizing compilers, visualization packages, performance analysis tools, etc.