Jump to content

Slurm Workload Manager

From Wikipedia, the free encyclopedia
Slurm
Developer(s)SchedMD
Stable release
Repository
Written inC
Operating systemLinux,BSDs
TypeJob Scheduler for Clusters and Supercomputers
LicenseGNU General Public License
Websiteslurm.schedmd

TheSlurm Workload Manager,formerly known asSimple Linux Utility for Resource Management(SLURM), or simplySlurm,is afree and open-sourcejob schedulerforLinuxandUnix-likekernels,used by many of the world'ssupercomputersandcomputer clusters.

It provides three key functions:

  • allocating exclusive and/or non-exclusive access to resources (computer nodes) to users for some duration of time so they can perform work,
  • providing a framework for starting, executing, and monitoring work, typically a parallel job such asMessage Passing Interface(MPI) on a set of allocated nodes, and
  • arbitrating contention for resources by managing a queue of pending jobs.

Slurm is the workload manager on about 60% of theTOP500supercomputers.[1]

Slurm uses abest fit algorithmbased onHilbert curve schedulingorfat treenetwork topology in order to optimize locality of task assignments on parallel computers.[2]

History[edit]

Slurm began development as a collaborative effort primarily byLawrence Livermore National Laboratory,SchedMD,[3]Linux NetworX,Hewlett-Packard,andGroupe Bullas a Free Software resource manager. It was inspired by the closed sourceQuadrics RMSand shares a similar syntax. The name is a reference to thesodainFuturama.[4]Over 100 people around the world have contributed to the project. It has since evolved into a sophisticated batch scheduler capable of satisfying the requirements of many large computer centers.

As of November 2021,TOP500list of most powerful computers in the world indicates that Slurm is the workload manager on more than half of the top ten systems.

Structure[edit]

Slurm's design is very modular with about 100 optional plugins. In its simplest configuration, it can be installed and configured in a couple of minutes. More sophisticated configurations provide database integration for accounting, management of resource limits and workload prioritization.

Features[edit]

Slurm features include:[citation needed]

  • No single point of failure, backup daemons, fault-tolerant job options
  • Highly scalable (schedules up to 100,000 independent jobs on the 100,000 sockets ofIBM Sequoia)
  • High performance (up to 1000 job submissions per second and 600 job executions per second)
  • Free and open-source software (GNU General Public License)
  • Highly configurable with about 100 plugins
  • Fair-share scheduling with hierarchical bank accounts
  • Preemptive and gang scheduling (time-slicing of parallel jobs)
  • Integrated with database for accounting and configuration
  • Resource allocations optimized for network topology and on-node topology (sockets, cores and hyperthreads)
  • Advanced reservation
  • Idle nodes can be powered down
  • Different operating systems can be booted for each job
  • Scheduling for generic resources (e.g.Graphics processing unit)
  • Real-time accounting down to the task level (identify specific tasks with high CPU or memory usage)
  • Resource limits by user or bank account
  • Accounting for power consumption by job
  • Support of IBM Parallel Environment (PE/POE)
  • Support for job arrays
  • Job profiling (periodic sampling of each task's CPU use, memory use, power consumption, network and file system use)
  • Sophisticated multifactor job prioritization algorithms
  • Support for MapReduce+
  • Support forburst bufferthat accelerates scientific data movement

The following features are announced for version 14.11 of Slurm, was released in November 2014:[5]

  • Improved job array data structure and scalability
  • Support for heterogeneous generic resources
  • Add user options to set the CPU governor
  • Automatic job requeue policy based on exit value
  • Report API use by user, type, count and time consumed
  • Communication gateway nodes improve scalability

Supported platforms[edit]

Slurm is primarily developed to work alongsideLinuxdistributions, although there is also support for a few otherPOSIX-basedoperating systems,includingBSDs(FreeBSD,NetBSDandOpenBSD).[6]Slurm also supports several unique computer architectures, including:

  • IBMBlueGene/Q models, including the 20 petaflopIBM Sequoia
  • CrayXT, XE and Cascade
  • Tianhe-2a 33.9 petaflop system with 32,000 Intel Ivy Bridge chips and 48,000 Intel Xeon Phi chips with a total of 3.1 million cores
  • IBM Parallel Environment
  • Anton

License[edit]

Slurm is available under theGNU General Public License v2.

Commercial support[edit]

In 2010, the developers of Slurm founded SchedMD, which maintains the canonical source, provides development, level 3 commercial support and training services. Commercial support is also available from Bull, Cray, and Science + Computing.

Usage[edit]

The `slurm` system has three main parts:

  • a central `slurmctld` (slurm control)daemonrunning on a single control node (optionally withfailoverbackups);
  • many computing nodes, each with one or more `slurmd` daemons;
  • clients that connect to the manager node, often withssh.

The clients can issue commands to the control daemon, which would accept and divide the workload to the computing daemons.

For clients, the main commands are `srun` (queue up an interactive job), `sbatch` (queue up a job), `squeue` (print the job queue), `scancel` (remove a job from the queue).

Jobs can be run inbatch modeorinteractive mode.For interactive mode, a compute node would start a shell, connects the client into it, and run the job. From there the user may observe and interact with the job while it is running. Usually, interactive jobs are used for initial debugging, and after debugging, the same job would be submitted by `sbatch`. For a batch mode job, its `stdout` and `stderr` outputs are typically directed to text files for later inspection.

See also[edit]

References[edit]

  1. ^"Running a Job on HPC using Slurm | HPC | USC".hpcc.usc.edu.Archived fromthe originalon 2019-03-06.Retrieved2019-03-05.
  2. ^Pascual, Jose Antonio; Navaridas, Javier; Miguel-Alonso, Jose (2009).Effects of Topology-Aware Allocation Policies on Scheduling Performance.Job Scheduling Strategies for Parallel Processing. Lecture Notes in Computer Science. Vol. 5798. pp. 138–144.doi:10.1007/978-3-642-04633-9_8.ISBN978-3-642-04632-2.
  3. ^"Slurm Commercial Support, Development, and Installation".SchedMD.Retrieved2014-02-23.
  4. ^"SLURM: Simple Linux Utility for Resource Management"(PDF).23 June 2003.Retrieved11 January2016.
  5. ^"Slurm - What's New".SchedMD.Retrieved2014-08-29.
  6. ^Slurm Platforms

Further reading[edit]

External links[edit]

Retrieved from "https://en.wikipedia.org/w/index.php?title=Slurm_Workload_Manager&oldid=1207033853"