lsf

LSF, short for Load Sharing Facility, is a commercial workload management and job-scheduling system for computer clusters and data centers. It was originally developed by Platform Computing and later acquired by IBM, where it is marketed as IBM Platform LSF. It is designed to coordinate the execution of large numbers of tasks across heterogeneous resources to maximize throughput and resource utilization.

Core capabilities include queueing and scheduling of jobs, resource management, dependencies between jobs, reservations, and accounting. It supports serial and parallel jobs, including MPI and OpenMP workloads, and can manage multi-queue environments, policy-driven priority, and fair-share scheduling. LSF can integrate with external inventory systems, cloud or grid resources, and provides monitoring and reporting for usage and performance.

Architecture typically features a centralized scheduler on a master host, a set of execution hosts on compute nodes, and client interfaces for submitting and monitoring jobs. Users submit work using commands such as bsub to enqueue jobs, bjobs to view status, bkill to cancel, and bqueues or bhosts to inspect queues and resources. The system can support high-availability configurations and scalable deployments across multiple clusters.

Applications span high-performance computing, scientific research, engineering simulations, and large-scale rendering for visual effects and media production. LSF's popularity stems from its mature ecosystem, enterprise features, and the ability to orchestrate resources across diverse environments.

LSF is an acronym with other meanings in different fields; in computing, however, it most commonly refers to the Load Sharing Facility.