Skip to content

Available batch job partitions

Note

The available LUMI batch job partitions are found in the LUMI documentation.

List all available partitions and their specifications on a system:

sinfo

Get details about a specific partition:

scontrol show partition <partition_name>

Puhti partitions

The following partitions (aka queues) are currently available in Puhti for normal (CPU) nodes:

Partition Time
limit		 Max
tasks		 Max
nodes		 Node types Max
memory		 Max
local storage
(NVMe)
test 15 minutes 80 2 M 185 GiB
interactive 7 days 8 1 IO 76 GiB 720 GiB
small 3 days 40 1 M, L, IO 373 GiB 3600 GiB
large 3 days 1040 26 M, L, IO 373 Gib 3600 GiB
longrun 14 days 40 1 M, L, IO 373 GiB 3600 GiB
hugemem 3 days 160 4 XL, BM 1496 GiB
hugemem_longrun 14 days 40 1 XL, BM 1496 GiB

The following partitions are available on GPU nodes. Note that for each GPU, you should reserve at most 10 cores/tasks.

Partition Time
limit		 Max
GPUs		 Max
nodes		 Node types Max
memory		 Max
local storage (NVMe)
gputest 15 minutes 8 2 GPU 373 GiB 3600 GiB
gpu 3 days 80 20 GPU 373 GiB 3600 GiB

Notes on partitions

  1. Use the test and gputest partitions for testing code
    • The available resources are limited, but the queuing times are short
  2. Only use the gputest and gpu partitions if you are sure your code uses GPUs
    • There are more CPU nodes than GPU nodes, so your job will run sooner if using CPUs
  3. Only request multiple CPU cores if you are sure your tool or code can use them
    • Reserving more cores does nothing by itself, except make you queue for longer
  4. Only request the memory you need
    • Memory may become a bottleneck for resource requests; requesting less memory will decrease your time in queue
    • Billing is based on memory requested, not on memory used
  5. Only use the longrun partitions if necessary
    • These partitions have stricter limits and a lower priority
  6. Only two jobs per user are allowed in the interactive partition
    • Use through apps in the web interface or with the command sinteractive

Mahti partitions

The following full-node partitions (aka queues) are currently available in Mahti on CPU nodes. When using these partitions, your jobs use all resources available on the node and won't share the node with other jobs.

Partition Nodes Time
limit		 Access
test 1-2 1 hours all
medium 1-20 36 hours all
large 20-200 36 hours scalability test
gc 200-700 36 hours Grand Challenge

The following partition is available for allocation by resources in Mahti on CPU nodes. This means that you can request a sub-node allocation: you can request only part of the resources (cores and memory) available on the compute node. This also means that your job may share the node with other jobs. Each core reserved will provide 1.875 GB of memory and the only way to increase the memory reservation is to increase the number of cores reserved.

Partition Cores Time
limit		 Max
memory		 Access
interactive 1-32 7 days 60 GiB all

The following partitions are available on GPU nodes. Note that for each full A100 GPU, you can reserve at most 32 cores/tasks and for each GPU the job is allocated 122.5 GiB of memory. A subset of the A100 GPUs are sliced into smaller a100_1g.5gb GPUs with one seventh of the compute and memory capacity of a full A100 GPU. For each small GPU you can reserve at most 4 cores/tasks and for each GPU the job is allocated 17.5 GiB of memory. Also note that you can reserve at most one small GPU per job.

Partition Time
limit		 Max
GPUs		 Max
nodes		 Node types Max
local storage (NVMe)
gputest 15 minutes 4 1 GPU 3800 GiB
gpusmall 36 hours 2 1 GPU 3800 GiB
gpumedium 36 hours 24 6 GPU 3800 GiB

Interactive partition, is intended for pre- and post processing tasks. It can be used both for interactive working and in batch jobs, where reserving a full node is not reasonable.

The basic CPU based HPC computing should be done in medium, large or gc partitions. Only full nodes are allocated from these partitions and the jobs will have access to all cores and memory in the nodes. You may choose to run with fewer cores if that gives better performance. In these partitions billing is based on the allocated nodes.

The large partition on Mahti is only accessible to projects that have done a scalability test and shown good utilization of the resources. The partition gc is only accessible to Grand Challenge projects and allows users to run extremely large simulations.