| Both sides previous revision
Previous revision
Next revision
|
Previous revision
|
equipment [2020/01/24 21:37]
root |
equipment [2022/02/02 17:31] (current)
jpummil |
| === Selecting Resources === | === Equipment/Selecting Resources === |
| We describe the resources available at AHPCC and how to select the best one for your computing job. | We describe the resources available at AHPCC and how to select the best one for your computing job. |
| Computing resources are presently divided into four clusters that use separate schedulers. This will be condensed in the future, as all logins will be moved to ''pinnacle.uark.edu'' and all schedulers are migrated to Slurm, with one or multiple Slurm schedulers to be determined. | Computing resources are presently divided into four clusters that use separate schedulers. This will be condensed in the future, as all logins will be moved to ''pinnacle.uark.edu'' and all schedulers are migrated to Slurm, with one or multiple Slurm schedulers to be determined. |
| |
| These efficient-use requirements do not apply to condo owners on their own nodes. | These efficient-use requirements do not apply to condo owners on their own nodes. |
| |
| == Trestles Cluster== | |
| |
| ''trestles'' includes about 200 public nodes with four AMD 6136 8-core 2.4 GHz processes, 64 GB of memory, and about 90 GB of temporary space in '/local_scratch/'. This generation of AMD processors has about half the per-core compute capacity of recent Intel processors, so the ''trestles'' 32-core nodes are roughly equivalent in computational power per-node in scalable programs to ''razor-2'' 16-core nodes below. However the ''trestles'' | |
| nodes have twice the memory. The ''trestles'' cluster also includes a number of Intel condo nodes. The ''trestles'' nodes are connected with slow but reasonably reliable Mellanox QDR Infiniband. | |
| |
| == Efficient and Best Use == | |
| ''trestles'' works well with scalable programs that fit into one node: 32 cores and 64 GB of memory. Multi-node programs on ''trestles'' will run better on ''pinnacle'', which in one node runs about the same as two to five ''trestles'' nodes, and has a much better network for larger jobs. Single-core to 12-core jobs will run better on ''razor-1'' or ''razor-2'' if the memory footprint of 24 or 32 GB works, as Intel cores are much faster. | |
| |
| If your job fits into one node of 32 cores and 64 GB, please use ''trestles'' to allow ''pinnacle'' to run larger jobs. | |
| |
| == Razor Clusters == | |
| The ''razor-1'' cluster has about 100 nodes with dual Intel X5670 processor, 12 cores, and 24 GB of memory. The ''razor-2'' cluster has about 80 nodes with dual Intel E5-2670v1 processor, 16 cores, and 32 GB of memory. The ''razor'' clusters have large local disks of about 900 GB capacity. They are connected with now troublesome QLogic QDR Infiniband. | |
| |
| == Efficient and Best Use == | |
| |
| The ''razor'' clusters also work well with programs that fit their core and memory footprint in one node, that is 12 cores and 24 GB on ''razor-1'' and 16 cores and 32 GB on ''razor-2''. The QLogic Infiniband on ''razor'' has become unreliable and multi-node jobs are not recommended. Jobs that require multiple nodes on ''razor'' are better run on ''pinnacle'' in one node or multiple nodes with EDR Infiniband. The ''razor'' clusters are also recommended for single core jobs. ''razor-1'' is generally less crowded if the single-core job uses less than 24 GB of memory. | |
| |
| == Overall Recommendations == | == Overall Recommendations == |
| We recommend the following clusters depending on the needs of your program and system load. | We recommend the following clusters depending on the needs of your program and system load. |
| These are rules of thumb not covering every possible situation, contact hpc-support@listserv.uark.edu with questions. | These are rules of thumb not covering every possible situation, contact hpc-support@listserv.uark.edu with questions. Here "memory" refers to shared memory of one node. |
| | |
| GPU-capable: use ''pinnacle'' GPU queues | |
| | |
| 1 to 12 cores and 24 GB memory: use ''razor-1'' | |
| | |
| 1 to 16 cores and 32 GB memory: use ''razor-2'' | |
| | |
| Up to 32 cores and 64 GB memory: use ''trestles'', though low-core count will be slow | |
| | |
| more than 64 GB shared memory: use ''pinnacle'' standard ''comp01/comp06/comp72'' | |
| | |
| more than 192 GB shared memory: use ''pinnacle'' ''himem06/himem72'' or high memory ''razor/trestles'' nodes | |
| |
| more than 32 cores: use ''pinnacle'' multiple nodes standard ''comp01/comp06/comp72'' | * GPU-capable |
| | * use ''pinnacle'' GPU queues |
| | * not GPU-capable |
| | * 1 to 12 cores and up to 24 GB memory: use ''razor-1'' |
| | * 1 to 16 cores and up to 32 GB memory: use ''razor-2'' |
| | * Up to 32 cores and up to 64 GB memory: use ''trestles'', though low-core count jobs will be slow compared with Intel |
| | * more than 64 GB shared memory, or all 32 cores: use ''pinnacle'' standard ''comp01/comp06/comp72'' |
| | * more than 192 GB shared memory: use ''pinnacle'' ''himem06/himem72'' or high memory ''razor/trestles'' nodes |
| | * more than 32 cores: use ''pinnacle'' multiple nodes standard ''comp01/comp06/comp72'' |
| |
| anything requiring four or more ''razor/trestles'' nodes: use ''pinnacle'' standard ''comp01/comp06/comp72'' | Discretionary cases: |
| | * anything requiring two or more ''razor/trestles'' nodes: ''pinnacle'' standard ''comp01/comp06/comp72'' will run much faster but probably start the job more slowly because of the job queue. |
| | * 1 node, 32 cores and less than 192 GB memory: use ''pinnacle'' standard or ''trestles'' if memory is less than 64 GB. ''pinnacle'' will run much faster but will probably start the job more slowly because of the job queue. |
| |
