Load Sharing Facility (LSF) is a set of software tools that distributes tasks across a cluster of workstations, balancing the load among them. There are a variety of utilities to submit tasks to the most lightly loaded machine within the cluster. Tasks may be explicitly submited to LSF using commands such as lsrun, or jobs can be submitted transparently from the lstcsh shell.
lstcsh is an LSF-enhanced version of tcsh. It includes a load-sharing capability with transparent remote job execution on the least loaded host in the cluster. Tasks are distributed and executed remotely based on a list of eligible commands.
see LSF Tasks and lsrtasks(1) for more info on lsf tasks.
lsmode flags control the verboseness of lstcsh in reporting execution details.
lsmode [-]e turn off/on display of remote command eligibility. [off]
lsmode [-]v turn off/on display of remote host executing command. [on]
lsmode [-]t turn off/on display of elapsed time report. [off]
lsmode local|remote controls whether command not in the remote/local command list get executed by default remotely or locally by default.
see lseligible(1) and lstcsh(1) for more info.
Use the lsrun command to explicitly submit jobs to the least loaded machine on the cluster.
For example:
tracy:-~/src/coredumpme >lsrun -v cc coredumpme.c <<Execute cc on remote host marple.umd.edu>> tracy:-~/src/coredumpme >
Use the lsload to check the load of the cluster or a specific machine.
tracy:-~ >lsload wolfe HOST_NAME status r15s r1m r15m ut pg ls it tmp swp mem wolfe.umd.edu ok 3.2 3.1 3.1 100% 0.0 14 0 466M 455M 63M tracy:-~ >lsload HOST_NAME status r15s r1m r15m ut pg ls it tmp swp mem holmes.umd.edu ok 1.5 1.4 1.5 94% 0.0 29 0 285M 452M 102M marple.umd.edu ok 1.6 1.5 1.4 100% 0.0 33 0 342M 427M 171M marlowe.umd.edu ok 1.9 1.6 1.7 100% 0.0 43 0 292M 366M 152M wolfe.umd.edu ok 3.1 3.1 3.1 100% 0.0 12 0 466M 455M 65M tracy.umd.edu ok 3.2 3.3 3.1 100% 0.0 40 0 292M 336M 67M tracy:-~ >