Milgram

Stanley

Milgram is named for Dr. Stanley Milgram, a psychologist who researched the behavioral motivations behind social awareness in individuals and obedience to authority figures. He conducted several famous experiments during his professorship at Yale University including the lost-letter experiment, the small-world experiment, and the Milgram experiment on obedience to authority figures.

Milgram is a HIPAA aligned Department of Psychology cluster intended for use on projects that may involve sensitive data. This applies to both storage and computation. If you have any questions about this policy, please contact us.

Note

Connections to Milgram can only be made from the Yale VPN (access.yale.edu)--even if you are already on campus (YaleSecure or ethernet). See our VPN page for setup instructions. If your group has a workstation (see list), you can connect using one of those.


Hardware

Milgram is made up of a couple kinds of compute nodes. The Features column below lists the features that can be used to request different node types using the --constraints flag (see our Slurm documentation for more details). The RAM listed below is the amount of memory available for jobs.

Warning

Care should be taken when scheduling your job if you are running programs/libraries optimized for specific hardware. You can narrow which nodes can run your job by requesting the features from the Node Configurations table as constraints (slurm --constraint flag) to your job. See the Request Compute Resources page and the Build Software page for further guidance.

Compute Node Configurations

Count CPU CPU Cores RAM GPU vRAM/GPU Features
12 2x E5-2660 v3 20 121G haswell, E5-2660_v3, nogpu, oldest
48 2x E5-2660 v4 28 250G broadwell, E5-2660_v4, nogpu
5 2x 6240 36 372G 4x rtx2080ti 8G cascadelake, avx512, 6240, nogpu, standard

Slurm Partitions

Nodes on the clusters are organized into partitions, to which you submit your jobs with Slurm. The default resource requests for all jobs is 1 core and 5GB of memory per core. The short partition is where most batch jobs should run, and is the default if you don't specify a partition. The interactive partition is dedicated to jobs with which you need ongoing interaction. The long and verylong partitions are meant for jobs with projected walltimes that are too long to run in short. For courses using the cluster we set aside the education partition. The scavenge partition allows you to run preemptable jobs on more resources than normally allowed. For more information about scavenge, see the Scavenge documentation.

The limits listed below are for all running jobs combined. Per-node limits are bound by the node types, as described in the hardware table.

Partition Group Limits User Limits Walltime Default/Max Node Type (count)
interactive 1 job, 4 CPUs, 20 G RAM 1h/6h E5-2660_v3 (2)
short* 1158 CPUs, 10176 G RAM 772 CPUs, 6784 G RAM 1h/6h E5-2660_v3 (9), E5-2660_v4 (48)
long 1188 CPUs, 5940 G RAM 1h/2d E5-2660_v3 (9), E5-2660_v4 (48)
verylong 792 CPUs, 3960 G RAM 1h/7d E5-2660_v3 (9), E5-2660_v4 (48)
gpu 1h/7d 6240 w/ rtx2080ti (5)
education 1h/6h E5-2660_v3 (2)
scavenge none E5-2660_v3 (9), E5-2660_v4 (48), 6240 (5)

* default

Storage

/gpfs/milgram is Milgram's primary filesystem where home, project, and scratch60 directories are located. For more details on the different storage spaces, see our Cluster Storage documentation.

You can check your current storage usage & limits by running the getquota command. Note that the per-user usage breakdown only update once daily.

Warning

Files stored in scratch60 are purged if they are older than 60 days. You will receive an email alert one week before they are deleted.

Partition Root Directory Storage File Count Backups
home /gpfs/milgram/home 20G/user 500,000 Yes
project /gpfs/milgram/project varies varies No
scratch60 /gpfs/milgram/scratch60 varies 5,000,000 No

Last update: March 2, 2020