Working on a single-node job scheduler for Linux. Large HPC clusters use schedulers like SLURM or PBS to manage allocation of resources to users, but these systems are quite overkill when all you have is a single node shared by a few users.
I am trying to offload as much of the complex stuff to existing parts of the kernel, like using systemd/cgroups for resource limiting and UNIX sockets for authentication.
I am trying to offload as much of the complex stuff to existing parts of the kernel, like using systemd/cgroups for resource limiting and UNIX sockets for authentication.