Job Submission Queues

From Research Computing Center Wiki
Jump to: navigation, search


Batch Queues on the Sapelo2

There are different queues defined on Sapelo2. Users are required to specify, in the job submission script or as job submission command line arguments, the queue and the resources needed by the job in order for it to be assigned to compute node(s) that have enough available resources (such as number of cores, amount of memory, GPU cards, etc).

Users can also specify the type of nodes to be used for a given job. The table below summarizes the queues defined and the compute nodes that they target:

Queue Name Node Type Description Notes
batch AMD 48-core, 128GB RAM, AMD Opteron, IB interconnect Regular nodes.
batch Intel 28-core, 64GB RAM, Intel Xeon, IB interconnect Regular nodes.
highmem_q HIGHMEM, AMD 48-core, 512GB RAM or 1TB, AMD Opteron, IB interconnect For high memory jobs.
highmem_q HIGHMEM, Intel 28-core, 1TB, Intel Xeon, IB interconnect For high memory jobs.
gpu_q GPU, K40 16-core, 128GB RAM, Intel Xeon, 8 NVIDIA K40m GPUs, IB interconnect For GPU-enabled jobs.
gpu_q GPU, K20 12-core, 96GB RAM, Intel Xeon, 7 NVIDIA K20Xm GPUs, IB interconnect For GPU-enabled jobs.

Each node with AMD Opteron processors has an AMD node feature. Similarly, each node with Intel Xeon processors has an Intel node feature. Jobs intended for an Intel HIGHMEM node can be submitted to the highmem_q and request a node with Intel node feature, as shown in the example below.

Nodes purchased by specific labs have different node features and are targeted by different queues. This information is provided separately to the users who have priority in using these buyin nodes.

For information on how to run jobs on Sapelo2, please see Running Jobs on Sapelo2.