Computational resources
Currently, we provide our users with 31640 computing cores, 183,50 TB RAM and disk space of more than 18 PB. The theoretical performance of this configuration is 1053 TFLOPS.
- 25 blade servers (HP-brand), each equipped with:
- 64 computing cores (4 processors AMD Opteron 6276);
- 256 GB of RAM (DDR3);
- 2 local 900 GB SAS hard disks;
- 4 network adapters (10-Gigabit Ethernet);
- 2 InfiniBand QDR ports (40 Gbit/s per port).
- The theoretical performance of single node: 588.8 GFLOPS.
- 180 blade servers (SuperMicro-brand), each equipped with:
- 40 computing cores (2 processors Intel Xeon E5-2680 v2);
- 128 GB of RAM (DDR3);
- Local hard disk Intel SSD 400 GB SATA;
- Network adapter 10-Gigabit Ethernet;
- InfiniBand QDR port (40 Gbit/s).
- The theoretical performance of single node: 448 GFLOPS.
- 144 blade servers (Bull-brand), each equipped with:
- 40 computing cores (2 processors Intel Xeon E5-2680 v2);
- 128 GB of RAM (DDR3);
- Local hard disk SSD 400 GB Micron P400m-MTFDDAK400MAN;
- Network adapter 1-Gigabit Ethernet;
- InfiniBand port 56 Gbit/s (together with InfiniBand to Ethernet 40 Gbit bridge).
- The theoretical performance of single node: 448 GFLOPS.
- Data storage (HP-brand):
- Data storage equipped with 140 hard disks 2TB SAS working as a high-performance distributed file system Lustre;
- The Lustre system consists of the MDS server (Metadata Server) along with 6 data servers OSS (Object Storage Server);
- SAS disks configured to operate in groups RAID-60 along with file system ZFS are assigned to a single MDT (Metadata Target) cooperating with the MDS server and 11 OST (Object Storage Target) cooperating with OSS servers;
- Communication between system servers and clients is conducted via the InfiniBand;
- The total capacity available to users: 170 TB;
- Read / write performance of the system reaches 2 GB/s.
- Data storage (NetApp-brand):
- 2 disk units containing a total amount of 328 3 TB SAS disks;
- 2 disk servers (model FAS6280) working in HA (High Availability) mode, each equipped with 96 GB of RAM, 512 GB cache Flash Cache and 4 Ethernet network adapters (10-Gigabit per port);
- File system: NFS version 4.1 with the extension pNFS;
- The total capacity available to users: 670 TB;
- Performance at the level of: 3 GB/s read and 1.5 GB/s write.
At the same time, for the needs of the experiments conducted at the European Organization for Nuclear Research CERN, we provide 320 computing cores, 1280 GB RAM and more than 236 TB of disk space.
- 5 blade servers (HP-brand), each equipped with:
- 64 computing cores (4 processors AMD Opteron 6276);
- 256 GB of RAM (DDR3);
- 2 local 900 GB SAS hard disks;
- 4 network adapters (10-Gigabit Ethernet);
- 2 InfiniBand QDR ports (40 Gbit/s per port);
- The theoretical performance of single node: 588.8 GFLOPS.
- Data storage (HP-brand):
- Data storage equipped with 140 hard disks 2TB SAS;
- SAS disks configured to operate in groups RAID-60 along with file system XFS;
- The total capacity available to users: 235 TB.
For all using our resources, we provide a fast and redundant Internet access. The connection to the external network is carried out by two independent circuits with capacity 10 gigabits per second.
Computing resources available to users are systematically expanded as new parts of equipment are delivered. Even distribution of purchases will not only enable the optimal use of financial resources in relation to the needs of users, but also provide opportunity to build the infrastructure basing on the latest solutions available on the market at the time.
The computer cluster built in the project is the one of the largest of this type in Poland and one of the most powerful supercomputers dedicated to the nuclear power industry in the world.