Working on a shared machine
Guide 4 · queues, limits and good habits
The Q-Chem server you use through IQmol is shared: by everyone in this class, by students in other courses, and by users around the world. Research clusters work the same way. Learning to use a shared resource well is one of the skills this lab teaches.
How a queue works
When you click Submit, your job doesn’t start right away. It joins a queue. A scheduler starts jobs as resources free up, usually first come, first served, sometimes with limits per user. Your job’s life:
queued → running → finished (or failed / killed)
The Job Monitor in IQmol shows which stage each job is in.
The limits of the free server
- Run time: each job can run for at most 10 minutes. After that it is killed and you get no energy.
- Concurrency: you may not be able to run many jobs at the same time. Extra jobs just wait in the queue.
- Availability: the server is a free service. It can be slow at busy times, and very occasionally unavailable. Never leave the calculations for the night before the deadline.
Every experiment page shows a rough cost flag next to each job: quick, a few minutes, or may hit the 10-min limit. These are estimates based on the molecule size, the basis set and the type of functional, not guarantees.
Good habits
- Test one job first. Before submitting 128 jobs, run one, open the output and check it.
- Submit in batches. For example, all the molecules for one functional, then check them, then the next functional.
- Name jobs systematically, e.g.
ex03_b3lyp_def2-tzvp_eth_st. The input generator does it for you. - Keep a job log. The checklist on each experiment page ticks itself off as you enter energies, so you can see at a glance what’s missing.
- Keep every output, organized in one folder per experiment.
- Don’t resubmit blindly. If a job fails, read the end of the output first. Resubmitting a broken input just wastes everyone’s queue time.
- Start early. Queues are longest right before deadlines, for every class, everywhere.
For these small molecules you could, but real research calculations need hundreds of cores and days of time. The workflow you learn here (prepare locally → submit to a queue → retrieve and analyze) is exactly how computational chemists work on university clusters and national supercomputers.