From PrimeHub v3.1, the new feature, Usage Reports, is introduced that administrators can have a overall insight of who/what consumed resources monthly.
{% hint style="info" %} Usage is defined by allocated resources, not by actual utilization. For example, when an user opens an Jupyter notebook, the record of the allocated resources is logged in the usage report, even if the user doesn't run any program actually on it. The each record includes the lifetime of a pod, and CPU/GPU/Memory are allocated/occupied for a pod. {% endhint %}
Click Detailed Report or Summary Report of the month for downloading a csv file or search specific year-month by the format YYYY/M (e.g. 2020/7, 2020/12) in Date search field.
You even can download the report of the current month which is not over yet. It will have a pop-up to inform you that the data of current month is not intact. Just click Confirm for downloading.
There are some insightful data of usage:
| Item | Description |
|---|---|
| report_month | the report is for which year and month |
| group | which group which component runs at |
| user | which user uses resources |
| component | such as job, notebook, model_deploy |
| component_name | the name of the component |
| cpu_core_hours | hours if the computing work runs in a CPU |
| gpu_core_hours | hours if the computing work runs in a GPU |
| gb_memory_hours | hours if the computing work uses 1 GB memory |
| usage_hours | hours the computing work has done |
| instance_type | instance type |
| instance_cpu_core | vCPU cores of the instance |
| instance_gpu_core | GPU cores of instance |
| instance_memory_gb | memory of the instance |
| pod_name | name of the pod |
| k8s_uid | Kubernetes object id |
| start_time | time pod began running |
| end_time | time pod finished running |
| running | if it's still running |
There are some insightful data of usage:
| Item | Description |
|---|---|
| report_month | the report is for which year and month |
| group | which group which component runs at |
| user | which user uses resources |
| component | such as job, notebook, model_deploy |
| gpu_core_hours | hours if the computing work runs in a GPU |
| cpu_core_hours | hours if the computing work runs in a CPU |
| gb_memory_hours | hours if the computing work uses 1 GB memory |
| usage_hours | hours the computing work has done |
| running | if it's still running |