When you first log into XDMoD, you will be directed to your Dashboard (to learn how yow to log into XDMoD, please view this document).
In order to view the job efficiency of any of your job, choose and click on the said job from "Job identifier" data and you will be directed to the "Job viewer" tab.
NOTE: If you are registered as a PI user on XDMoD, you will see your jobs along with the jobs of your entire group.
An alternate way to do this is directly navigating to the "Job Viewer" tab and clicking on "Search" in the top left corner of the page
Now, here you can search for your job in two ways. If you know the Job Id for your job, the easiest way to look it up is "Quick Job Lookup". Alternatively, the "Advanced search" method can be used to find jobs based on multiple parameters.As mentioned earlier the easiest and quickest way to look up a job on "Job Viewer" is Quick Job Lookup, so much so you wouldn't have to use the Advanced search method at all.
To do that, you would just need your Job ID, which if you don't remember, can be looked up when you're connected to one of our clusters and run the command "slogs". Pick the appropriate option for Resource (ub hpc in most cases), and click on search.
Once you complete any one of the steps mentioned above, you will be able to view performance data on your finished job.
Here, you will the see values "CPU User", "Homogeneity", "CPU User Balance", and "Memory Headroom".
In a gist:
"CPU User": This is the ratio of your CPU time to the total CPU time for the cores that was assigned to the job. The higher the number, the better the job performed.
"Homogeneity": It indicates how uniform the L1D load rate was throughout the run time of your job. The closer the number rating is to 1, the better.
"CPU User Balance": It indicates how well balanced the CPU usage is between the various cores that the job was assigned. A value of 1 indicates that all the cores were working equally, while a value of 0 indicates that one or more nodes were idle.
"Memory Headroom": This gives a measure of amount of memory used for that job. It is recommended to not request more/extra memory unless its required for the job. Hence, being sure about how much memory you're going to need for a specific job before running it can prove to be beneficial. Here, the closer the number is to 1, the better.
You must have noticed that there are more data available for the job you selected:
The "Accounting data" tab shows information about the job such as the start and end time, little information on user, and the account that was charged.
The "Job Script" as you might have guessed, displays the script of the job.
"Executable Information" shows information about the processes that were run on the compute nodes during the job.
"Summary metrics" displays more detailed data on the performance metrics that it collected when your job was running.
"Detailed metrics": shows the data generated by the job summarization software.
For more information on these available data, please view this page.
To learn how to edit charts, please view this document.
To learn how to use Metric explorer, please view this document.
To learn how to export and download charts fro XDMoD, please view this document.