Greeting all HPC Users,
In the past few months, we noticed that many jobs have been pending for too long due to the 14 days maximum wall time per job. This means that if the previous jobs are to run until finish, all the pending jobs will need to wait for a maximum of 14 days before they can start to execute.
After internal discussion, we have come to a conclusion that the maximum wall time per job should be reduced to a more reasonable time. Seeing there were many small jobs in the past, we decided to reduce the maximum wall time per job to a maximum of 7 days for QoS long. This should help clear up the queue faster and reduce maximum waiting time per job in the long run, so user jobs should only need to wait at most half of the original waiting time compared to the previous setting.
Meanwhile, we also encourage you to utilize a full node if your application can do so. This should help speed up your calculation, so you can obtain your research result faster. This also means jobs that used to take 14 days with half a node, might take less than 7 days to complete.
If you have any concern or questions, do let us know through the service desk.
Thank you.