|
4 gadi atpakaļ | |
---|---|---|
asciinema | 4 gadi atpakaļ | |
images | 4 gadi atpakaļ | |
README.md | 4 gadi atpakaļ |
First of all, it's worth checking if your supervisor already has accesss to HPC services and then requesting an account there.
If not you can request an account:
This guide is going to use the CS HPC cluster but the two are very similar. Worth noting that I have borrowed loads from the much better docs of RC https://www.rc.ucl.ac.uk/docs/.
The IHI is in the process so procuring its own HPC services but we're not there yet :(.
Most people use something like the following workflow:
You will need to either use the UCL Virtual Private Network or ssh in to UCL's gateway tails.ucl.ac.uk
first. From tails you can then ssh in.
ssh <your_UCL_user_id>@tails.cs.ucl.ac.uk
ssh <your_UCL_user_id>@<login_node>.cs.ucl.ac.uk
There are a few login nodes availabel but it shouldn't really matter which you use. gamble
is the one I use.
You can setup a
Submit a job to the scheduler with qsub
qsub /path/to/submission/script/
Get the status of a job with qstat
qstat
job-ID prior name user state submit/start at queue slots ja-task-ID
-----------------------------------------------------------------------------------------------------------------
6506636 0.00000 testing jbloggs qw 21/12/2012 11:11:11 1
qstat -j <job-ID>
Submit a job to the scheduler with qsub
qsub /path/to/submission/script/