Difference between revisions of "Nexus"

From UMIACS
Jump to navigation Jump to search
(30 intermediate revisions by the same user not shown)
Line 1: Line 1:
The Nexus is the combined scheduler of resources in UMIACS.  Many of our existing computational clusters that have discrete schedulers will be folding into this scheduler in the future.  The resource manager for Nexus (as with our other existing computational clusters) is [[SLURM]].  Resources are arranged into partitions where users are able to schedule computational jobs.  Users are arranged into a number of Slurm accounts based on faculty, lab, or center investments.
+
The Nexus is the combined scheduler of resources in UMIACS.  Many of our existing computational clusters that have discrete schedulers will be folding into this scheduler in the future.  The resource manager for Nexus (as with our other existing computational clusters) is [[SLURM]].  Resources are arranged into partitions where users are able to schedule computational jobs.  Users are arranged into a number of SLURM accounts based on faculty, lab, or center investments.
  
 
= Getting Started =
 
= Getting Started =
Line 5: Line 5:
  
 
== Access ==
 
== Access ==
The submission nodes for the Nexus computational resources are determined by department, center or lab affiliation.  You can log into the UMIACS Directory application and select their [https://intranet.umiacs.umd.edu/directory/cr/ Computational Resources] (CR).  They will find a CR that has the prefix <code>nexus</code>, select it and in the Host section will list the available login nodes.
+
The submission nodes for the Nexus computational resources are determined by department, center, or lab affiliation.  You can log into the [https://intranet.umiacs.umd.edu/directory/cr/ UMIACS Directory CR application] and select the Computational Resource (CR) in the list that has the prefix <code>nexus</code>. The Hosts section lists your available login nodes.
  
 
'''Note''' - UMIACS requires multi-factor authentication through our [[Duo]] instance.  This is completely discrete from both UMD's and CSD's Duo instances.  You will need to enroll one or more devices to access resources in UMIACS, and will be prompted to enroll when you log into the Directory application for the first time.
 
'''Note''' - UMIACS requires multi-factor authentication through our [[Duo]] instance.  This is completely discrete from both UMD's and CSD's Duo instances.  You will need to enroll one or more devices to access resources in UMIACS, and will be prompted to enroll when you log into the Directory application for the first time.
  
Once you have identified your submission nodes, you are able to [[SSH]] directly into them.  From there, you are able to submit to the cluster via our [[SLURM]] workload manager.  You need to make sure that your submitted jobs have the correct account, partition, and qos.
+
Once you have identified your submission nodes, you can [[SSH]] directly into them.  From there, you are able to submit to the cluster via our [[SLURM]] workload manager.  You need to make sure that your submitted jobs have the correct account, partition, and qos.
  
 
== Jobs ==
 
== Jobs ==
[[SLURM]] jobs are submitted by either <code>srun</code> or <code>sbatch</code> depending if you are doing an interactive job or batch job, respectively.  You need to provide the where/who the job will run and specify the resources you need to run with. There are defaults for both, so if you don't specify something you may be scheduled with a very minimal set of time and resources (including NO GPUs unless specifically requested). 
+
[[SLURM]] jobs are submitted by either <code>srun</code> or <code>sbatch</code> depending if you are doing an interactive job or batch job, respectively.  You need to provide the where/how/who to run the job and specify the resources you need to run with.
  
For the where/who, you may be required to specify <code>--account</code>, <code>--qos</code> and/or <code>--partition</code> to be able to adequately submit jobs to the Nexus.
+
For the where/how/who, you may be required to specify <code>--partition</code>, <code>--qos</code>, and/or <code>--account</code> (respectively) to be able to adequately submit jobs to the Nexus.
  
For resources, you may need to specify for cpus (<code>--tasks</code>), memory (<code>--mem</code>), and GPUs (<code>--gres=gpu</code>) in your submission arguments to meet your requirements.  For more information about submitting for GPU resources, see [[SLURM/JobSubmission#Requesting_GPUs]].  You can also can run <code>man srun</code> on your submission node for a complete list of available submission arguments.
+
For resources, you may need to specify <code>--time</code> for time, <code>--tasks</code> for CPUs, <code>--mem</code> for RAM, and <code>--gres=gpu</code> for GPUs in your submission arguments to meet your requirements.  There are defaults for all four, so if you don't specify something, you may be scheduled with a very minimal set of time and resources (e.g., by default, NO GPUs are included if you do not specify <code>--gres=gpu</code>).  For more information about submission flags for GPU resources, see [[SLURM/JobSubmission#Requesting_GPUs]].  You can also can run <code>man srun</code> on your submission node for a complete list of available submission arguments.
  
 
=== Interactive ===
 
=== Interactive ===
Line 32: Line 32:
 
The SLURM resource manager uses partitions to act as job queues which can restrict size, time and user limits.  The Nexus (when fully operational) will have a number of different partitions of resources.  Different Centers, Labs, and Faculty will be able to invest in computational resources that will be restricted to approved users through these partitions.
 
The SLURM resource manager uses partitions to act as job queues which can restrict size, time and user limits.  The Nexus (when fully operational) will have a number of different partitions of resources.  Different Centers, Labs, and Faculty will be able to invest in computational resources that will be restricted to approved users through these partitions.
 
* [[Nexus/Tron]] - This is the pool of resources available to all UMIACS and CSD faculty and graduate students.  It provides access for undergraduate and graduate teaching resources.
 
* [[Nexus/Tron]] - This is the pool of resources available to all UMIACS and CSD faculty and graduate students.  It provides access for undergraduate and graduate teaching resources.
* Scavenger - This is a [https://slurm.schedmd.com/preempt.html preemption] partition that supports nodes from multiple other partitions.  Jobs are subject to preemption rules however more resources are available to schedule simultaneously than in other partitions.  You are responsible for ensuring your jobs handle this preemption correctly, as the SLURM scheduler will simply restart each job with the same submission arguments when preempted jobs are available to run again.
+
* Scavenger - This is a [https://slurm.schedmd.com/preempt.html preemption] partition that supports nodes from multiple other partitions.  More resources are available to schedule simultaneously than in other partitions, however jobs are subject to preemption rules.  You are responsible for ensuring your jobs handle this preemption correctly, as the SLURM scheduler will simply restart each job with the same submission arguments when preempted jobs are available to run again.
  
 
= Quality of Service (QoS) =
 
= Quality of Service (QoS) =
SLURM uses a QoS to provide limits on job sizes to users.  Note that you should still try to only allocate the minimum resources for your jobs, as resources that each of your jobs schedules are counted against your FairShare priority in the future.
+
SLURM uses a QoS to provide limits on job sizes to users.  Note that you should still try to only allocate the minimum resources for your jobs, as resources that each of your jobs schedules are counted against your [https://slurm.schedmd.com/fair_tree.html FairShare priority] in the future.
 +
* default - Default QoS. Limited to 4 cores, 32GB RAM, and 1 GPU per job.  The maximum wall time per job is 3 days.  4 jobs are permitted simultaneously.
 +
* medium - Limited to 8 cores, 64GB RAM, and 2 GPUs per job .  The maximum wall time per job is 2 days.  2 jobs are permitted simultaneously.
 +
* high - Limited to 16 cores, 128GB RAM, and 4 GPUs per job.  The maximum wall time per job is 1 day.  Only 1 job is permitted simultaneously.
 +
* scavenger - Limited to 64 cores, 256GB RAM, and 8 GPUs per job.  The maximum wall time per job is 2 days.  Only 16 GPUs are permitted simultaneously.  This QoS is both only available in the scavenger partition and the only QoS available in the scavenger partition. To use this QoS, include <code>--partition=scavenger</code> and <code>--account=scavenger</code> in your submission arguments. Do not include any QoS argument other than <code>--qos=scavenger</code> or the submission will fail.
  
* default - Default QoS. Limited to 4 cores, 32GB RAM and 1 GPU. The maximum wall time per job is 3 days and 4 jobs are permitted simultaneously.
+
You can display these QoSes from the command line using <code>show_qos</code> command. Other lab-or-group-specific QoSes or reserved QoSes may also appear in the listing. The above four QoSes are the ones that everyone can submit to.
* medium - Limited to 8 cores, 64GB RAM, 2 GPUs.  The maximum wall time per job is 2 days and 2 jobs are permitted simultaneously.
 
* high - Limited to 16 cores, 128GB RAM, 4 GPUs.  The maximum wall time per job is 1 day and only 1 job is permitted simultaneously.
 
* scavenger - Limited to 64 cores, 256GB RAM and 8 GPUs. The maximum wall time per job is 2 days and users will be able to run up to 16 GPUs simultaneously.  This QoS is only available in the scavenger partition.
 
  
You can display these QoSes from the command line using <code>show_qos</code> command.
 
 
<pre>
 
<pre>
 
# show_qos
 
# show_qos
      Name    MaxWall MaxJobs                        MaxTRES    MaxTRESPU  Priority
+
            Name    MaxWall MaxJobs                        MaxTRES    MaxTRESPU  Priority
---------- ----------- ------- ------------------------------ ------------- ----------
+
---------------- ----------- ------- ------------------------------ ------------- ----------
scavenger  2-00:00:00            cpu=64,gres/gpu=8,mem=256G  gres/gpu=16          0
+
          normal                                                                          0
    medium  2-00:00:00      2      cpu=8,gres/gpu=2,mem=64G                        0
+
      scavenger  2-00:00:00            cpu=64,gres/gpu=8,mem=256G  gres/gpu=16          0
      high  1-00:00:00      1    cpu=16,gres/gpu=4,mem=128G                        0
+
          medium  2-00:00:00      2      cpu=8,gres/gpu=2,mem=64G                        0
  default  3-00:00:00      4      cpu=4,gres/gpu=1,mem=32G                        0
+
            high  1-00:00:00      1    cpu=16,gres/gpu=4,mem=128G                        0
      tron                                                      gres/gpu=4          0
+
        default  3-00:00:00      4      cpu=4,gres/gpu=1,mem=32G                        0
 +
            tron                                                      gres/gpu=4          0
 +
      gamma-long 10-00:00:00            cpu=32,gres/gpu=8,mem=256G                        0
 
</pre>
 
</pre>
  
Line 59: Line 61:
  
 
= Storage =
 
= Storage =
All storage available in Nexus is currently NFS based.  We will be introducing some changes for Phase 2 to support high performance GPUDirect Storage (GDS).  These storage allocation procedures will be revised and approved by the launch of Phase 2 by a CSD and UMIACS faculty committee.
+
All storage available in Nexus is currently [[NFS]] based.  We will be introducing some changes for Phase 2 to support high performance GPUDirect Storage (GDS).  These storage allocation procedures will be revised and approved by the launch of Phase 2 by a joint UMIACS and CSD faculty committee.
  
 
== Home Directories ==
 
== Home Directories ==
Each user account in UMIACS is allocated 20GB of storage in their home directory (/nfshomes/$username).  This file system has snapshots and backups availableThe quota is fixed however and is not available to increase.
+
Home directories in the Nexus computational infrastructure are available from the Institute's [[NFShomes]] as <code>/nfshomes/USERNAME</code> where USERNAME is your username.  These home directories have very limited storage (20GB, cannot be increased) and are intended for your personal files, configuration and source codeYour home directory is '''not''' intended for data sets or other large scale data holdingsUsers are encouraged to utilize our [[GitLab]] infrastructure to host your code repositories.
 +
 
 +
'''NOTE''': To check your quota on this directory you will need to use the <code>quota -s</code> command.
 +
 
 +
Your home directory data is fully protected and has both [[Snapshots | snapshots]] and is [[NightlyBackups | backed up nightly]].
  
 
In Phase 2, other standalone compute clusters will fold into partitions in Nexus and you will start to have the same home directory across all systems.
 
In Phase 2, other standalone compute clusters will fold into partitions in Nexus and you will start to have the same home directory across all systems.
  
 
== Scratch Directories ==
 
== Scratch Directories ==
Each user will be allocated a 200GB scratch directory under <code>/fs/nexus-scratch/$username</code>.  If your directory is completely filled, you may request a permanent increase of up to 400GB total.  This space is does not have snapshots and is not backed up.  Please ensure that any data you have under your scratch directory is reproducible.
+
Scratch data has no data protection including no snapshots and the data is not backed up. There are two types of scratch directories in the Nexus compute infrastructure:
 +
* Network scratch directory
 +
* Local scratch directories
 +
 
 +
=== Network Scratch Directory ===
 +
You are allocated 200GB of scratch space via NFS from <code>/fs/nexus-scratch/$username</code>.  '''It is not backed up or protected in any way.'''  This directory is '''automounted''' so you will need to <code>cd</code> into the directory or request/specify a fully qualified file path to access this.
 +
 
 +
You may request a permanent increase of up to 400GB total space without any faculty approval by [[HelpDesk | contacting staff]]If you need space beyond 400GB, you will need faculty approval and/or a project directory.
 +
 
 +
This file system is available on all submission, data management, and computational nodes within the cluster.
 +
 
 +
=== Local Scratch Directories ===
 +
Each computational node that a user can schedule compute jobs on also has one or more local scratch directories.  These are always named <code>/scratch0</code>, <code>/scratch1</code>, etc.  These are almost always more performant than any other storage available to the job.  However, users must stage their data within the confine of their job and stage the data out before the end of their job.
 +
 
 +
These local scratch directories have a tmpwatch job which will '''delete unaccessed data after 90 days''', scheduled via maintenance jobs to run once a month at 1am.  Different nodes will run the maintenance jobs on different days of the month to ensure the cluster is still highly available at all times.  Please make sure you secure any data you write to these directories at the end of your job.
  
 
== Faculty Allocations ==
 
== Faculty Allocations ==
Each faculty will have 1TB of lab space to be allocated to them when their account is installed.  We also can support grouping these individual allocations together into larger center, lab, or research group allocations if desired by the faculty.  Please contact staff@umiacs.umd.edu to inquire.
+
Each faculty member can be allocated 1TB of lab space upon request.  We can also support grouping these individual allocations together into larger center, lab, or research group allocations if desired by the faculty.  Please [[HelpDesk | contact staff]] to inquire.
  
This lab space will by default not have snapshots (but are available if requested) and it is backed up.
+
This lab space does not have [[Snapshots | snapshots]] by default (but are available if requested), but is [[NightlyBackups | backed up]].
  
 
== Project Allocations ==
 
== Project Allocations ==
 
Project allocations are available per user for 270 TB days; you can have a 1TB allocation for up to 270 days, a 3TB allocation for 90 days, etc..  A single faculty member can not have more than 20 TB of sponsored account project allocations active at any point.  
 
Project allocations are available per user for 270 TB days; you can have a 1TB allocation for up to 270 days, a 3TB allocation for 90 days, etc..  A single faculty member can not have more than 20 TB of sponsored account project allocations active at any point.  
  
To request an allocation, please send mail to staff@umiacs.umd.edu with your account sponsor CC'd.  Please include the following details:
+
The minimum storage space you can request (maximum length) is 500GB (540 days) and the minimum allocation length you can request (maximum storage) is 30 days (9TB).
  
 +
To request an allocation, please [[HelpDesk | contact staff]] with your account sponsor involved in the conversation.  Please include the following details:
 
* Project Name (short)
 
* Project Name (short)
 
* Description
 
* Description
Line 84: Line 105:
 
* Length in days (270 days, 135 days, etc.)
 
* Length in days (270 days, 135 days, etc.)
  
These allocations will be available via <code>/fs/nexus-projects/$project_name</code>.
+
These allocations will be available via <code>/fs/nexus-projects/$project_name</code>. Near the end of the allocation period, staff will contact you and ask you if you want to renew for the same duration; you cannot exceed the original 270 TB days limit for renewals.
  
 
== Datasets ==
 
== Datasets ==
Datasets are hosted in <code>/fs/nexus-datasets</code>.  If you want to request a dataset for for consideration, please email staff@umiacs.umd.edu. We will have a more formal process to approve datasets by phase 2 of Nexus.  Please note that datasets that require accepting a license will need to be reviewed by [https://ora.umd.edu/ UMD's Office of Research Administration (ORA)] which may require some time to process.
+
We have read-only dataset storage available at <code>/fs/nexus-datasets</code>.  If there are datasets that you would like to see curated and available, please see [[Datasets | this page]].
 +
 
 +
We will have a more formal process to approve datasets by phase 2 of Nexus.

Revision as of 13:32, 10 June 2022

The Nexus is the combined scheduler of resources in UMIACS. Many of our existing computational clusters that have discrete schedulers will be folding into this scheduler in the future. The resource manager for Nexus (as with our other existing computational clusters) is SLURM. Resources are arranged into partitions where users are able to schedule computational jobs. Users are arranged into a number of SLURM accounts based on faculty, lab, or center investments.

Getting Started

All accounts in UMIACS are sponsored. If you don't already have a UMIACS account, please see Nexus/Accounts for information on getting one.

Access

The submission nodes for the Nexus computational resources are determined by department, center, or lab affiliation. You can log into the UMIACS Directory CR application and select the Computational Resource (CR) in the list that has the prefix nexus. The Hosts section lists your available login nodes.

Note - UMIACS requires multi-factor authentication through our Duo instance. This is completely discrete from both UMD's and CSD's Duo instances. You will need to enroll one or more devices to access resources in UMIACS, and will be prompted to enroll when you log into the Directory application for the first time.

Once you have identified your submission nodes, you can SSH directly into them. From there, you are able to submit to the cluster via our SLURM workload manager. You need to make sure that your submitted jobs have the correct account, partition, and qos.

Jobs

SLURM jobs are submitted by either srun or sbatch depending if you are doing an interactive job or batch job, respectively. You need to provide the where/how/who to run the job and specify the resources you need to run with.

For the where/how/who, you may be required to specify --partition, --qos, and/or --account (respectively) to be able to adequately submit jobs to the Nexus.

For resources, you may need to specify --time for time, --tasks for CPUs, --mem for RAM, and --gres=gpu for GPUs in your submission arguments to meet your requirements. There are defaults for all four, so if you don't specify something, you may be scheduled with a very minimal set of time and resources (e.g., by default, NO GPUs are included if you do not specify --gres=gpu). For more information about submission flags for GPU resources, see SLURM/JobSubmission#Requesting_GPUs. You can also can run man srun on your submission node for a complete list of available submission arguments.

Interactive

Once logged into a submission node, you can run simple interactive jobs. If your session is interrupted from the submission node, the job will be killed. As such, we encourage use of a terminal multiplexer such as Tmux.

$ srun --pty --ntasks 4 --mem=2gb --gres=gpu:1 nvidia-smi -L
GPU 0: NVIDIA RTX A4000 (UUID: GPU-ae5dc1f5-c266-5b9f-58d5-7976e62b3ca1)

Batch

Batch jobs are scheduled with a script file with an optional ability to embed job scheduling parameters via variables that are defined by #SBATCH lines at the top of the file. You can find some examples in our SLURM/JobSubmission documentation.

Partitions

The SLURM resource manager uses partitions to act as job queues which can restrict size, time and user limits. The Nexus (when fully operational) will have a number of different partitions of resources. Different Centers, Labs, and Faculty will be able to invest in computational resources that will be restricted to approved users through these partitions.

  • Nexus/Tron - This is the pool of resources available to all UMIACS and CSD faculty and graduate students. It provides access for undergraduate and graduate teaching resources.
  • Scavenger - This is a preemption partition that supports nodes from multiple other partitions. More resources are available to schedule simultaneously than in other partitions, however jobs are subject to preemption rules. You are responsible for ensuring your jobs handle this preemption correctly, as the SLURM scheduler will simply restart each job with the same submission arguments when preempted jobs are available to run again.

Quality of Service (QoS)

SLURM uses a QoS to provide limits on job sizes to users. Note that you should still try to only allocate the minimum resources for your jobs, as resources that each of your jobs schedules are counted against your FairShare priority in the future.

  • default - Default QoS. Limited to 4 cores, 32GB RAM, and 1 GPU per job. The maximum wall time per job is 3 days. 4 jobs are permitted simultaneously.
  • medium - Limited to 8 cores, 64GB RAM, and 2 GPUs per job . The maximum wall time per job is 2 days. 2 jobs are permitted simultaneously.
  • high - Limited to 16 cores, 128GB RAM, and 4 GPUs per job. The maximum wall time per job is 1 day. Only 1 job is permitted simultaneously.
  • scavenger - Limited to 64 cores, 256GB RAM, and 8 GPUs per job. The maximum wall time per job is 2 days. Only 16 GPUs are permitted simultaneously. This QoS is both only available in the scavenger partition and the only QoS available in the scavenger partition. To use this QoS, include --partition=scavenger and --account=scavenger in your submission arguments. Do not include any QoS argument other than --qos=scavenger or the submission will fail.

You can display these QoSes from the command line using show_qos command. Other lab-or-group-specific QoSes or reserved QoSes may also appear in the listing. The above four QoSes are the ones that everyone can submit to.

# show_qos
            Name     MaxWall MaxJobs                        MaxTRES     MaxTRESPU   Priority
---------------- ----------- ------- ------------------------------ ------------- ----------
          normal                                                                           0
       scavenger  2-00:00:00             cpu=64,gres/gpu=8,mem=256G   gres/gpu=16          0
          medium  2-00:00:00       2       cpu=8,gres/gpu=2,mem=64G                        0
            high  1-00:00:00       1     cpu=16,gres/gpu=4,mem=128G                        0
         default  3-00:00:00       4       cpu=4,gres/gpu=1,mem=32G                        0
            tron                                                       gres/gpu=4          0
      gamma-long 10-00:00:00             cpu=32,gres/gpu=8,mem=256G                        0

Currently in our non-preemption partition, you will be restricted to 4 GPUs at once.

To find out what accounts and partitions you have access to, use the show_assoc command.

Storage

All storage available in Nexus is currently NFS based. We will be introducing some changes for Phase 2 to support high performance GPUDirect Storage (GDS). These storage allocation procedures will be revised and approved by the launch of Phase 2 by a joint UMIACS and CSD faculty committee.

Home Directories

Home directories in the Nexus computational infrastructure are available from the Institute's NFShomes as /nfshomes/USERNAME where USERNAME is your username. These home directories have very limited storage (20GB, cannot be increased) and are intended for your personal files, configuration and source code. Your home directory is not intended for data sets or other large scale data holdings. Users are encouraged to utilize our GitLab infrastructure to host your code repositories.

NOTE: To check your quota on this directory you will need to use the quota -s command.

Your home directory data is fully protected and has both snapshots and is backed up nightly.

In Phase 2, other standalone compute clusters will fold into partitions in Nexus and you will start to have the same home directory across all systems.

Scratch Directories

Scratch data has no data protection including no snapshots and the data is not backed up. There are two types of scratch directories in the Nexus compute infrastructure:

  • Network scratch directory
  • Local scratch directories

Network Scratch Directory

You are allocated 200GB of scratch space via NFS from /fs/nexus-scratch/$username. It is not backed up or protected in any way. This directory is automounted so you will need to cd into the directory or request/specify a fully qualified file path to access this.

You may request a permanent increase of up to 400GB total space without any faculty approval by contacting staff. If you need space beyond 400GB, you will need faculty approval and/or a project directory.

This file system is available on all submission, data management, and computational nodes within the cluster.

Local Scratch Directories

Each computational node that a user can schedule compute jobs on also has one or more local scratch directories. These are always named /scratch0, /scratch1, etc. These are almost always more performant than any other storage available to the job. However, users must stage their data within the confine of their job and stage the data out before the end of their job.

These local scratch directories have a tmpwatch job which will delete unaccessed data after 90 days, scheduled via maintenance jobs to run once a month at 1am. Different nodes will run the maintenance jobs on different days of the month to ensure the cluster is still highly available at all times. Please make sure you secure any data you write to these directories at the end of your job.

Faculty Allocations

Each faculty member can be allocated 1TB of lab space upon request. We can also support grouping these individual allocations together into larger center, lab, or research group allocations if desired by the faculty. Please contact staff to inquire.

This lab space does not have snapshots by default (but are available if requested), but is backed up.

Project Allocations

Project allocations are available per user for 270 TB days; you can have a 1TB allocation for up to 270 days, a 3TB allocation for 90 days, etc.. A single faculty member can not have more than 20 TB of sponsored account project allocations active at any point.

The minimum storage space you can request (maximum length) is 500GB (540 days) and the minimum allocation length you can request (maximum storage) is 30 days (9TB).

To request an allocation, please contact staff with your account sponsor involved in the conversation. Please include the following details:

  • Project Name (short)
  • Description
  • Size (1TB, 2TB, etc.)
  • Length in days (270 days, 135 days, etc.)

These allocations will be available via /fs/nexus-projects/$project_name. Near the end of the allocation period, staff will contact you and ask you if you want to renew for the same duration; you cannot exceed the original 270 TB days limit for renewals.

Datasets

We have read-only dataset storage available at /fs/nexus-datasets. If there are datasets that you would like to see curated and available, please see this page.

We will have a more formal process to approve datasets by phase 2 of Nexus.