In the past, allocated resources were requested, granted, and accounted in normalized core hours (NCH). NCH used a factor different for each cluster and type of nodes.
With progressing time, the system of NCH has become obsolete, since we now have clusters that greatly vary in computing capacity, i.e. CPU, GPU, memory, and determining the value of the resources of different clusters is a complex process that cannot be based on a single value.
As a result, the accounting policy has been changed from normalized core hours to node hours.
This means that in the future, it is required to apply for node hours of the specific cluster and node type.
You can specify the number of node hours requested for each platform:
a) Barbora CPU
https://docs.it4i.cz/barbora/compute-nodes/#compute-nodes-without-accelerators
b) Barbora GPU
https://docs.it4i.cz/barbora/compute-nodes/#compute-nodes-with-a-gpu-accelerator
c) Barbora FAT
https://docs.it4i.cz/barbora/compute-nodes/#fat-compute-node
d) DGX-2
https://docs.it4i.cz/dgx2/introduction/
e) Karolina CPU
https://docs.it4i.cz/karolina/compute-nodes/#compute-nodes-without-accelerators
f) Karolina GPU
https://docs.it4i.cz/karolina/compute-nodes/#compute-nodes-with-a-gpu-accelerator
g) Karolina FAT
https://docs.it4i.cz/karolina/compute-nodes/#data-analytics-compute-node
If you have any questions, please contact support@it4i.cz