Hardware : Différence entre versions
Ligne 92 : | Ligne 92 : | ||
<div class="alert"> | <div class="alert"> | ||
− | This paragraph describes the obsolete "legacy nef" configuration which will definitively stop on 17 april 2016 : please use the [[User_Guide_new_config| new nef configuration]]. | + | This paragraph describes the obsolete "legacy nef" configuration which will definitively stop on 17 april 2016 : please use the [[User_Guide_new_config| new nef configuration]]. |
+ | |||
+ | Hardware from the legacy cluster will be re-installed in the new cluster | ||
</div> | </div> | ||
Ligne 219 : | Ligne 221 : | ||
* common storage : /home | * common storage : /home | ||
** '''15 TiB''' (8 x 4TB SAS disks, 2 x RAID-10 array), infiniband QDR, NFS access, available to all users, quotas | ** '''15 TiB''' (8 x 4TB SAS disks, 2 x RAID-10 array), infiniband QDR, NFS access, available to all users, quotas | ||
− | |||
* capacity distributed and scalable common storage : /data | * capacity distributed and scalable common storage : /data | ||
** '''153TiB''' , infiniband QDR, BeeGFS access | ** '''153TiB''' , infiniband QDR, BeeGFS access | ||
Ligne 226 : | Ligne 227 : | ||
** scratch storage : variable size (initially ~20TiB), no quota limit, for transient storage (data may be purged) | ** scratch storage : variable size (initially ~20TiB), no quota limit, for transient storage (data may be purged) | ||
+ | Legacy team storage : | ||
* legacy external storage ([http://www.dell.com/fr/entreprise/p/powervault-md1200/pd Dell MD1200]) : out of maintenance hardware | * legacy external storage ([http://www.dell.com/fr/entreprise/p/powervault-md1200/pd Dell MD1200]) : out of maintenance hardware | ||
** 19TB for the [https://team.inria.fr/asclepios ASCLEPIOS] EPI | ** 19TB for the [https://team.inria.fr/asclepios ASCLEPIOS] EPI | ||
Ligne 231 : | Ligne 233 : | ||
* legacy external storage ([http://www.dell.com/downloads/global/products/pvaul/fr/pvaul_md1000_specs_fr.pdf Dell MD1000]) : out of maintenance hardware | * legacy external storage ([http://www.dell.com/downloads/global/products/pvaul/fr/pvaul_md1000_specs_fr.pdf Dell MD1000]) : out of maintenance hardware | ||
** 1TB RAID-1 for the [http://www-sop.inria.fr/teams/tropics TROPICS] EPI archive | ** 1TB RAID-1 for the [http://www-sop.inria.fr/teams/tropics TROPICS] EPI archive | ||
+ | |||
= legacy cluster - Storage = | = legacy cluster - Storage = |
Version du 16 mars 2016 à 17:45
Sommaire
new cluster - CPU & GPU
The current platform includes:
- 8 dual-Xeon 8 cores, 16GB/core RAM
- 44 dual-Xeon 6 cores, 8 GB/core RAM
- 13 quad-Xeon 6 cores, ~2.6 GB/core RAM
The number of cores available is
- 128 Xeon 2.6Ghz cores : ~2.4 Tflops (2.7 Tflops peak)
- 528 Xeon 2.93Ghz cores : ~4.8 Tflops (6.2 Tflops peak)
- 312 Xeon 2.4Ghz cores : ~1.8 Tflops (3.0 Tflops peak)
The total cumulated computing power is close to 11.9 Tflops (peak, double precision)
Dell C6220 Cluster (2015)
Dell C6220 dual-Xeon E5-2650 v2 @ 2.60GHz (16 cores) : 8 nodes ( nef029 to nef036 )
- RAM capacity : 256 GB RAM
- 1x1TB SATA HardDisk drive
- 2x gigabit network ports (one connected)
- 2x infiniband QDR card (one connected)
Dell C6100 Cluster
Dell C6100 dual-Xeon X5670 @ 2.93GHz (12 cores) : 44 nodes ( nef097 to nef140 )
- RAM capacity : 96 GB RAM
- 1x250GB hard disk drive SATA 7.2kRPM
- 2x gigabit network ports (one connected)
- 1x infiniband QDR card
Dell R900 Cluster
Dell R900 quad-Xeon E7450 @ 2.4GHz (24 cores) : 13 nodes ( nef084 to nef096 )
- RAM capacity : 64 GB RAM
- 2x146GB hard disk drive (RAID-1) SAS 15K RPM
- 4x gigabit network ports (one connected)
- 1x infiniband QDR card
Dell C6100/C410x GPU cluster
Dell C6100 nodes: mono-Xeon X5650 @ 2.66Ghz : 2 nodes ( nefgpu05 nefgpu06)
- RAM capacity: 24 GB
- 1x250GB SATA disk
- 2x gigabit ethernet ports (one connected)
- 1x infiniband QDR card
- 2 Tesla M2050 GPUs connected with a PCIe gen2 16x interface
- 448 Streaming Processor Cores per card
- Simple precision performance peak: 1.03 Tflops per card
- Double precision performance peak: 0.51 Tflops per card
- 3GB of RAM capacity per card
Carri GPU cluster
Carri HighStation 5600 XLR8 nodes: dual-Xeon X5650 @ 2.66Ghz : 1 node ( nefgpu03 )
- RAM capacity: 72 GB
- 2x160GB SSD disks
- 4x gigabit ethernet ports (one connected)
- 1x infiniband QDR card
- 7 Tesla C2070 (nefgpu03)/ 7 Tesla C2050(nefgpu04) GPUs connected with a PCIe gen2 16x interface
- 448 Streaming Processor Cores per card
- Simple precision performance peak: 1.03 Tflops per card
- Double precision performance peak: 0.51 Tflops per card
- 3GB of RAM capacity per card (6GB on nefgpu03)
legacy cluster - CPU & GPU
This paragraph describes the obsolete "legacy nef" configuration which will definitively stop on 17 april 2016 : please use the new nef configuration.
Hardware from the legacy cluster will be re-installed in the new cluster
The current platform includes:
- 6 quad-Opteron 12 cores
- 6 quad-Opteron 16 cores
- 16 dual-Xeon 10 cores.
- 21 dual-Xeon quad core.
- 2 dual-Xeon hexa core.
- 2 mono-Xeon hexa core.
- 22 Nvidia Tesla GPU
The number of cores available is
- 288 Opteron 2.2Ghz cores : ~2 Tflops (2.53 Tflops peak)
- 384 Opteron 2.3Ghz cores : ~2.7 Tflops (3.5 Tflops peak)
- 320 Xeon E5 2.8Ghz cores : ~5.9 Tflops (7.17 Tflops peak)
- 204 Xeon 2.66Ghz cores : ~1.7 Tflops (2.17 Tflops peak)
- 9024 Streaming Processor Cores (22 Nvidia GPU) ~9.5Tflops peak (22.2 Tflops in single precision)
The total cumulated computing power is close to 24 Tflops (peak, double precision)
Dell R815 Cluster
Dell R815 quad-Opterons 6174 @ 2.2Ghz (48 cores) : 6 nodes
- RAM capacity : 256 GB RAM
- 2x600GB SAS HardDisk drive (RAID-0)
- 4x gigabit network ports (one connected)
- 1x infiniband QDR card
Dell C6145 Cluster
Dell C6145 quad-Opterons 6376 @ 2.3Ghz (64 cores) : 6 nodes
- RAM capacity : 256 GB RAM (512GB on nef011 and nef012)
- 1x500GB SATA HardDisk drive
- 2x gigabit network ports (one connected)
- 1x infiniband QDR card
Dell C6220 Cluster (2014)
Dell C6220 dual-Xeon E5-2680 v2 @ 2.80GHz (20 cores) : 16 nodes
- RAM capacity : 192 GB RAM
- 1x2TB SATA HardDisk drive
- 2x gigabit network ports (one connected)
- 1x infiniband FDR card (QDR used)
Dell 1950 Cluster
PowerEdge 1950 dual-Xeon 5355 @ 2.66Ghz : 19 nodes
- RAM capacity : 16 GB
- 2x73GB SAS in RAID-0, 146Go disk space available
- 2x gigabit netword ports (one connected)
- 1x infiniband QDR card
HP GPU cluster
HP DL160 G5 nodes: dual-Xeon 5430 @ 2.66Ghz : 2 nodes
- RAM capacity: 16 GB
- 500GB of local disk space
- 2x gigabit ethernet interface
- 1x 10GBit myrinet interface
- 2 GPUs connected with a PCIe gen2 16x interface
Nvidia Tesla S1070: 1 node
- 4 GPUs (2 per compute node)
- Streaming Processor Cores: 960 (260 per GPU)
- Simple precision performance peak: 3.73 Tflops
- Double precision performance peak: 0.31 Tflops
- RAM capacity : 16 GB (4GB per GPU)
Carri GPU cluster
Carri HighStation 5600 XLR8 nodes: dual-Xeon X5650 @ 2.66Ghz : 1 node
- RAM capacity: 72 GB
- 2x160GB SSD disks
- 4x gigabit ethernet ports (one connected)
- 1x infiniband QDR card
- 7 Tesla C2070 (nefgpu03)/ 7 Tesla C2050(nefgpu04) GPUs connected with a PCIe gen2 16x interface
- 448 Streaming Processor Cores per card
- Simple precision performance peak: 1.03 Tflops per card
- Double precision performance peak: 0.51 Tflops per card
- 3GB of RAM capacity per card (6GB on nefgpu03)
new cluster - Storage
All nodes have access to common storage :
- common storage : /home
- 15 TiB (8 x 4TB SAS disks, 2 x RAID-10 array), infiniband QDR, NFS access, available to all users, quotas
- capacity distributed and scalable common storage : /data
- 153TiB , infiniband QDR, BeeGFS access
- to be extended : + 55 TiB in early 2016
- permanent storage : 1TiB quota per team + teams may buy additional quota (please contact cluster administrators)
- scratch storage : variable size (initially ~20TiB), no quota limit, for transient storage (data may be purged)
Legacy team storage :
- legacy external storage (Dell MD1200) : out of maintenance hardware
- legacy external storage (Dell MD1000) : out of maintenance hardware
- 1TB RAID-1 for the TROPICS EPI archive
legacy cluster - Storage
This paragraph describes the obsolete "legacy nef" configuration which will definitively stop on 17 april 2016 : please use the new nef configuration
All nodes have access to a common storage using NFS. The NFS server is accessed through the infiniband 40Gb network if available
- common storage : /home
- 15 TBytes (8 x 4TB SAS disks, 2 x RAID-10 array), NFS access, available to all users, quotas
- experimental scratch storage based on distributed file system : /dfs
- 19 TB, GlusterFS access, available to all users, no quotas (may change later)
- legacy external storage (Dell MD1200) : out of maintenance hardware
- legacy external storage (Dell MD1000) : out of maintenance hardware
- 1TB RAID-1 for the TROPICS EPI archive