Skip to content



Job accounting on Lucia is based on the actual resources mobilized by the jobs, taking requested CPUs, memory, and GPUs (when applicable) into account. For instance, a job on a standard node requesting only 1 CPU along with all the memory available on the node will be charged the full node.

Moreover, as there are various types of nodes, resources are normalized using weights, with 1 CPU of a "standard" compute node as the base billing unit, see table below.

The resource usage is then measured in billing unit hour (BUh) that can be calculated as follows:

  • For jobs running in shared mode or ExclusiveUser mode:

    MAX(req_cpu, req_mem_gb * (cpu_pnode / mem_pnode_gb), [req_gpu * (cpu_pnode / gpu_pnode)]) * weight * job_duration_h

  • For jobs running in exclusive mode:

    num_nodes * num_cpu_pnode * weight * job_duration_h

The table below shows each type of node with their corresponding partition, weight, billing unit hour for the minimal amount of resource that can be requested, and billing unit hour for 1 full node:

Node type Partition Weight Min BUh BUh for 1 full node
Standard (cns) batch, debug 1 1 (1 CPU) 128
Medium (cnm) medium 1.25 1.25 (1 CPU) 160
Large (cnl) large 4 4 (1 CPU) 256
Extra-large (cnx) xlarge 5 5 (1 CPU) 320
GPU (cna) gpu 16 128 (1 GPU) 512
AI (cni) ia 25.6 204.8 (1 GPU) 1638.4
Visualization (cng) visu 4 32 (1 GPU) 128

Credit management

Credits in BUh are allocated to each project, users can check the credits allocated to their projects and the corresponding usage with the get_slurm_quota command. The command also displays the expiration date of the project.

Projects that exhausted their BUh credits or passed their expiration date will be automatically blocked for new job submission, therefore project managers are kindly requested to monitor their project resources usage. Academic users can always update their project credits (and other information such as disk quota limits) via the following page: Other users can submit a service request via


Weekly and monthly usage reports are not implemented yet, in the meantime, usage for a given time period can be checked with the sreport command. Here's an example requesting the usage for the project "my_project" for the month of August 2023:

sreport cluster AccountUtilizationByUser Accounts=my_project format=account,login,used -t Hour -T billing Start=2023-08-01 End=2023-09-01