Hi, You prob need to do some investigation on what HPC users requirements are. For instance, genomics researchers need lots of cores and memory. For those users a 16 vCPU, 64GB RAM won't make the grade to be called HPC. They would need to have multiple vCPUs and use MPI to distribute their workload, and that adds complexity to their code. HPC nodes that I'm running in a cluster each have 64 cores and a minimum of 768 GB RAM each and some nodes have 1500 GB RAM. Even then some of the genomics jobs run for a few days to several hundred hours.
The other problem is that users would probably not know how to write a docker file to create their custom compute environment.
It would be a nice project though :-)
Mike
No comments yet.