HPC Dashboard
A desktop app that shows what your HPC jobs actually used — not what you booked.
Your scheduler tells you a job finished. It does not tell you that it used 37% of the CPU you paid for, or that 30 of your 48 cores sat idle for six hours. That waste is invisible unless something measures it while the job runs.
Two pieces
| logger-rs | Runs on the HPC alongside your job and records what it used. Writes plain files to shared storage. |
| This dashboard | Runs on your computer, reads those files over SFTP or a local mount, and turns them into charts and recommendations. |
Nothing runs on the HPC on your behalf. The dashboard only reads files. It never submits, cancels or modifies a job, and there is no service to install on the cluster.
Start here
| Page | |
|---|---|
| Installing | Download, first run, and getting past the OS security warning |
| Collecting Telemetry | Add logger-rs to your job script and mark the workload to measure |
| Connecting to an HPC | SSH keys and setting up a profile |
Using it
| Page | |
|---|---|
| Reading a Job | What each tab shows and how to interpret it |
| Advice and Recommendations | What to request next time, and why two obvious savings aren't real |
| Comparing Jobs | Two runs side by side |
When something's wrong
| Page | |
|---|---|
| Troubleshooting | Connection failures, empty job lists, missing charts |
| FAQ | Privacy, cost, what it does and does not do |
The quickest possible start
- Install the app — download, open, dismiss one security warning.
- Add telemetry to your job script and submit it.
- Add a profile pointing at where those logs land.
- In Jobs, choose View charts (or Download for charts first for a large log), then open the Cores tab.
Step 2 is the one people skip. Without it there are no telemetry files, and the dashboard will show an empty job list however correct your profile is.