LPJS | Lightweight , Portable Job Scheduler

LPJS is a C library. LPJS has no bugs, it has no vulnerabilities, it has a Permissive License and it has low support. You can download it from GitHub.

LPJS (Lightweight Portable Job Scheduler) is a batch system, i.e. a job scheduler and resource manager for HPC (High Performance Computing) clusters. An HPC cluster is anywhere from one to thousands of computers (called nodes) with managed CPU and memory resources for the purpose of performing computationally intensive jobs. Most clusters are linked together with a private high-speed network to maximize the speed of data exchange between the nodes. Typically there is a "head node", dedicated to keeping track of available CPU cores and memory on all nodes, multiple compute nodes, and one or more file servers, A.K.A. I/O nodes. Using LPJS and similar systems, users can queue jobs to run as soon as resources become available. Compute nodes with available cores and memory are automatically selected and programs usually run unattended (called batch mode), redirecting terminal output to files. It is possible to run interactive jobs as well, but this is rare and typically only used for debugging. Jobs may start running as soon as they are submitted, or they may wait in the queue until sufficient resources become available. Either way, once you have submitted a job, you can focus on other things, knowing that your job will complete as soon as possible. Users should, however, keep a close eye on their running jobs to make sure they are working properly. This avoids wasting resources and shows common courtesy to other cluster users. Unlike other batch systems, LPJS is designed to be small, easy to deploy and manage, and portable to any POSIX platform. Most existing batch systems are extremely complex, including our long-time favorite, SLURM, which stands for "Simple Linux Utility for Resource Management", but is no longer simple by any stretch of the imagination. The 'S' in SLURM has become somewhat of an irony as it has evolved into the premier batch system for massive and complex clusters. Note that THERE IS NOTHING INHERENTLY COMPLICATED ABOUT AN HPC CLUSTER. In its typical basic form, it's just a LAN with a manager node, a file server and some software to manage computing resources. You can make a cluster as complicated as you wish, but simple HPC clusters can be highly effective at reducing computation time by orders of magnitude. Overly complex HPC tools present a barrier to learning and research for those who have no ready access to centralized HPC resources. In major research institutions, centralizing HPC into large clusters can improve utilization of resources and reduce overall costs. In many organizations, however, building a large cluster and staffing a support group is simply not feasible. Large HPC clusters are dominated by Redhat Enterprise Linux (RHEL) and its derivatives for good reasons. For one thing, RHEL is the only platform besides Windows supported by many commercial science and engineering applications such as ANSYS, Fluent, Abacus, etc. Unfortunately, RHEL achieves enterprise reliability and long-term binary compatibility by using older, time-tested and debugged Linux kernels, compilers, and other tools, which often make it difficult to run the latest open source software. Facilitating small-scale HPC on platforms other than RHEL can eliminate this issue for open source software users.