Inula
Inula is an AMD and GPGPU cluster. It has been acquired from POWIEW and PL-Grid projects resources.
Hardware characteristics | ||
---|---|---|
Type | PC cluster | PC cluster |
Architecture | AMD (Interlagos) | Intel Xeon E5-2697 |
Network interfaces | Gigabit Ethernet 1000/10000 | InfiniBand FDR |
Network interconnections | InfiniBand QDR | |
Number of CPU cores | 3360 | 3584 |
Total computing power | 138,4 TFLOPS | 165 TFLOPS |
System memory | 6,56 TB | 12,3 TB |
Disk memory | 43,17 TB | |
Number of GPU modules | 205 |
Node merchant name | CPU model | No. of nodes | No. of processors and cores | RAM per node | Computing power | GPGPU model | Node name | Node class tag |
---|---|---|---|---|---|---|---|---|
SGI/RACKABLE | AMD Opteron™ 6234 | 70 | 2x12 | 64 GB | 57,6 GFLOPS | 2 x TeslaM2050 | c001-c021 | cane |
SGI/RACKABLE | AMD Opteron™ 6234 | 70 | 2x12 | 48 GB | 57,6 GFLOPS | 1 x TeslaM2050 | c001-c021 | cane |
HUAWEI CH121 V3 | Intel Xeon E5-2697 v3 | 65 | 2x14 | 64 GB | 1,1 TFLOPS | h1-h62;h136-h138 | huawei | |
HUAWEI CH121 V3 | Intel Xeon E5-2697 v3 | 73 | 2x14 | 128 GB | 1,1 TFLOPS | h63-h135 | huawei | |
HP ProLiant BL2x220c G6 | Intel Xeon 5530 | 292 | 2x4 | 16 GB | 22425.6 GFLOPS | r148-r164, r273-r399 | reef | |
HP ProLiant BL465c G7 | AMD Opteron(tm) Processor 6164 HE | 69 | 2x12 | 11260.8 GFLOPS | ||||
Iceotope | Intel Xeon E5-2620 | 42 | 2x6 | 32 GB | 192 GFLOPS | k01-k40 | kujawski |
Other information | ||
---|---|---|
System | ||
Type | GNU Linux | |
Computer domain name | inula.man.poznan.pl | |
Task management system in the batch mode | Portable Batch System (PBS) |
Queues
Launching jobs via PBS is performed using queues with varying priorities, limits or access rights (even resources available)
We have the following queues:
Name
Kolejki |
Job time limit | Default job time limit | Standard RAM capacity | Notes |
---|---|---|---|---|
multi | 7 dni | 1 dzień | 2 GB | This queue is for jobs running simultaneously on several nodes (e.g.. applications using MPI library) |
standard | 7 days | 1 day | 2 GB | This queue is for jobs running on a single node. Maximum value for ppn is 8 (ppn=8) |
plgrid-long | 7 days | 3 days | - | |
plgrid | 3 days | - | - | |
fast | 1 hour | 10 minutes | - | high priority; maximum no. of nodes: 1 |
interactive | 1 hour | 24 hours | - | maximum no. of nodes: 1 |