Job scheduler

Molecular simulation is computationally intensive. Therefore, it is usually impractical to run simulations on local computer. Instead, simulation jobs are usually submitted to a job scheduler on a distributed computing infrastructure. Then the jobs will be spread and executed on the compute nodes.

Wrappers for several job schedulers are provided in mstk. They are designed to simplify the most common operations of job schedulers, like querying, submitting and deleting jobs, generate job scripts, etc… It also provide a cache to the scheduler so that when high-throughput simulations are performed, the job scheduler will not experience much pressure.

PbsJob(id, name, state[, workdir, user, queue])	A PbsJob object is a job managed by scheduler.
JobParameter([queue, n_proc, n_gpu, n_node, ...])	The parameters for submitting a job
Scheduler()	Base class for job schedulers.
Slurm()	Slurm job scheduler with support of GPU allocation and MPI/OpenMP hybrid parallelization.
RemoteSlurm(host, username, remote_dir[, port])	Slurm job scheduler running on a remote machine.