Check usage costs for current fiscal year

It is useful to take a look at usage over periods of time and calculate overall usage at osc. This page explains how to do this using the HPC Job Activity tool in my.osc.edu

After logging into the client portal, navigate to Individual -> HPC Job Activity.

Enter the appropriate dates:

min date: 01 July 2019
max date: current date (or other end date)

Once the dates are entered, the click the button Refresh Report.

It will then produce a table with all the job records within that timeframe.

CPU usage costs

Next, a new column can be created using existing job information. This column will use the core hours that a job used times the cost per core hour of $0.003, giving the cost of the job.

Actions -> Data -> Compute:
label: base cpu cost
Compute Expression: AW * 0.003

AW is the variable corresponding to core hours of a job record. A list of the variables will be visible on the Compute screen.
On the same window as above, one can optionally specify a format mask. If it should represent a dollar amount, then enter the text:
FML999G999G999G999G990D0000
as the format mask and it will add a $ symbol to the result.

Since there are different prices associated with cpu, gpu and hugememory node usage, filter by the appropriate charge type.

Actions -> Filter:

ChargeType contains CPU

Now there should only be job records dealing with normal compute nodes, along with the cost of those jobs.

Finally, these jobs can be grouped together using different criteria. Select Actions -> Group by:
Project

Also, use the functions option to sum the core hours and base cpu cost of every job grouped by Project:
Sum: core hours
Sum: **base cpu cost

If there are multiple projects, then select the checkbox to the right of the function line to sum the results of each project.

Again, the sum of base cpu cost can be formatted to add a $ symbol like above using the text:
FML999G999G999G999G990D0000
Custom columns will have ** in front of the label chosen when selecting them from the function option of Group by.

The results will show the sum of costs for the period specified, separated by project code. However, this is only the cpu hours cost, not including separate gpu and huge memory node costs.

Make sure to save the report just created!
Actions -> Report -> Save Report
To delete a report, select it, then click the X next to Saved Report = <report-name>

Huge Memory usage costs

The same process can be repeated for calculating huge memory node usage costs. Edit the report that is currently open from above, but save it as a different name at the end.

The cost of using a huge memory node is slightly higher than normal comput nodes, so create a new computation.

The Group by filter must be removed to add a new computation.

Actions -> Data -> Compute:
label: hugemem cost
Compute Expression: AW * 0.004

Change the filter so that only huge memory job records are shown.

Actions -> Filter:

ChargeType contains HUGEMEM

Now, perform a similar Group by as with the previous instructions.

Actions -> Group by:
Project

Functions:
Sum core hours
Sum **hugemem cost

Again, save this report with a unique name so that these steps do not need to be repeated!

GPU usage costs

Finally, gpu costs can be estimated in the same way.

Remember, the Group by filter must be removed to add a new computation.

Actions -> Data -> Compute:
label: gpu cost
Compute Expression: AW * 0.045

Change the filter so that only gpu job records are shown.

Actions -> Filter:

ChargeType contains GPU

Finally, use Group by like before.

Actions -> Group by:
Project

Functions:
Sum core hours
Sum **gpu cost

SAVE THE REPORT!!!

Export reports

The reports created above can also be saved locally as a csv. Select Actions -> Download and click the available format to download.

Service: