Billing¶

Running jobs and storing data on the parallel filesystem consume the billing units allocated to your project:

for computing, your project is allocated CPU-core-hours
for storage, your project is allocated GB-hours.

Compute billing¶

For compute, your project is allocated CPU-core-hours that are consumed when running jobs. Depending on the partition, the way this billing is carried out differs.

Standard partition¶

The standard partition is operated in exclusive mode: the entire node will always be allocated. In practice, you 128 core-hours are billed for every allocated node and per hour even if your job requested less than 128 cores per node.

For example, 16 nodes for 12 hours:

16 nodes x 12 hours x 128 core-hour = 24576 core-hours

Small partition¶

When using the small partition you are billed per allocated core or if you are above a certain threshold per chunk of 2GB of memory. Here is the formula that is used for billing:

corehours = max(ncore, ceil(mem/2GB)) x time

if you use less than 2GB of memory per core, you are charged per allocated cores
if you use more than 2GB of memory per core, you are charged per 2GB slice of memory
if you are using the large memory nodes you will be billed per 2GB slice of memory

For example, 4 cores, 4GB of memory for 1 day:

4 cores x 24 hours = 96 core-hours

For example, 4 cores, 32GB of memory for 1 day:

32GB / 2GB x 24 hours = 384 core-hours

Checking core hours used¶

You can use the commands sreport and sacct to see how many core hours your project has consumed so far. These commands query the accounting database used by SLURM and are always up to date. For example, to get a summary of a how much a certain project has run start from the start of a certain date up until now, you can write:

sreport -t hours cluster AccountUtilization account=project_465000XXX start=2022-01-01 end=now

Example output:

--------------------------------------------------------------------------------
  Cluster         Account     Login     Proper Name       Used   Energy 
--------- --------------- --------- --------------- ---------- -------- 
     lumi project_465000+                               739228        0 
     lumi project_465000+     spock           Spock     120228        0 
     lumi project_465000+      data            Data     300000        0 
     lumi project_465000+      tpol           T'Pol     319000        0

The top row is summary for all project members. Please note that SLURM counts usage in CPU thread hours, so the numbers need to be divided by 2 to get the corresponding CPU core hours.

Storage billing¶

Storage is billed by volume as well as time. The billing units are GB-hours.

Regular Lustre filesystem¶

On the regular (spinning disk) Lustre filesystem, 1GB of data consume 1GB-hour out of your storage allocation for every hour it stays on the filesystem.

For example, 375GB for 4 days:

375 GB x 4 days x 24 hours = 36000 GB-hours

Flash Lustre filesystem¶

The flash based filesytem is billed at a 10x rate: 1GB of data consume 10GB-hour out of your storage allocation for every hour it stays on the filesystem. As a consequence, if you don't want to consume your storage allocation too quickly, it's recommended to remove your data from the flash filesystem as soon as possible.

For example, 150GB for 2 days:

150 GB x 2 days x 24 hours x 10 = 72000 GB-hours