Snowy User Guide

This is the user guide to Snowy, a high performance computer cluster at UPPMAX. Guides for the other systems at UPPMAX can be found here.

Please read this Users Guide for up-to-date information.


All heavy usage of the cluster must go through the batch system, SLURM

System configuration

Snowy consists of 228 compute servers (nodes) where each compute server consists of two 8-core Xeon E5-2660 processors running at 2.2 GHz. We provide 198 nodes with 128 GB memoryi (s1-s120, s151-s228) , 13 nodes with 256 GB (s138-s150) and 17 nodes with 512 GB (s121-s137). All nodes are interconnected with a 2:1 oversubscribed FDR (40 GB/s) Infiniband fabric. In total Snowy provides 3548 CPU cores in compute nodes.

OS and software

There are several compilers available through the module system on Snowy. This gives you flexibility to obtain programs that run optimally on Snowy.

  • gcc - the newest version usually generates the best code, if you tell it to use the new instructions. Check which version is the newest by doing module avail.
    The compiler executable is named gcc for C, g++ for C++, and gfortran for Fortran.
    To use the new instructions available on Snowy (AVX2 and FMA3), give the additional options "-mavx2 -mfma3" to gcc. For good performance with this compiler you should also specify optimization at least at level -O2 or -O3. Also try using -march=broadwell for GCC >= 4.9.0 or -march=core-avx2 for GCC 4.8.x, which will enable all the instructions on the CPU.
  • Intel+MKL - usually generates the fastest code. As with gcc, it is good to use the latest version. The compiler executable is named icc for C, icpc for C++, and ifort for Fortran. You should give optimization options at least -O2, preferably -O3 or -fast. You can also try to use the -xCORE-AVX2 option to the compiler to output AVX2 instructions.
  • pgi - often generates somewhat slower code, but it is stable so often it is easier to obtain working code, even with quite advanced optimizations. The compiler executable is named pgcc for C, pgCC for C++, and pgfortran, pgf77, pgf90, or pgf95 for Fortran. For this compiler, you can generate code for Snowy using the following options "UPDATES NEEDED". Also give optimization options at least -O2, preferably -Ofast, even though the compile times are much longer, the result is often worth the wait.

See the our software pages for more details about OS, compilers and installed software.

You will probably have good use of the following commands:

  • uquota - telling you about your file system usage.
  • projinfo - telling you about the CPU hour usage of your projects.
  • jobinfo - telling you about running and waiting jobs on Snowy.
  • finishedjobinfo - telling you about finished jobs on Snowy.
  • projmembers - telling you about project memberships.
  • projsummary [project id] - summarizes some useful information about projects

For SLURM commands and for commands like projinfo, jobinfo and finishedjobinfo, you may use the "-M" flag to ask for the answer to be given for a system that you are not logged in to. E.g., when logged into Rackham, you may ask about information about current core hour usage on Snowy, with the command projinfo -M Snowy

Accounts and log in

Snowy is different from other clusters at UPPMAX in that there are no login nodes for Snowy. All access to this system is done via secure shell (a.k.a SSH) interactive login to the login node, using the domain name rackham.uppmax.uu.se

ssh -AX user@rackham.uppmax.uu.se

To get a user account and start using UPPMAX, see the Getting Started page.

For questions concerning accounts and access to Rackham and Snowy, please contact UPPMAX support.

Note that the machine you arrive at when logged in is only a so called login node, where you can do various smaller tasks. We have some limits in place that restricts your usage. For larger tasks you should use our batch system that pushes your jobs onto other machines within the cluster.

All access to Snowy is done using the batch system SLURM, either as an interactive job or non-interactive batch jobs.

Using the batch system

To allow a fair and efficient usage of the system we use a resource manager to coordinate user demands. On Snowy we use the SLURM resource manager. Read our SLURM user guide for detailed information on how to use SLURM.

Note: When acessing snowy from Rackhams login nodes you must always use the flag -M for all SLURM commands.
Examples:

  • squeue -M snowy
  • jobinfo -M snowy
  • sbatch -M snowy slurm_script_file
  • scancel -u username -M snowy
  • interactive -A projectname -M snowy -p node -n 32 -t 01:00:00

Note: We always recommend loading all your modules in your job script file, This is even more important when running on Snowy since the module environment is not the same on the Rackham login nodes as on Snowy compute nodes.

Some Limits

  • There is a job walltime limit of 30 days (720 hours).
  • We restrict each user to at most 5000 running and waiting jobs in total.
  • Each project has a 30 days running allocation of CPU hours. We do not forbid running jobs after the allocation is overdrafted, but instead allow to submit jobs with a very low queue priority, so that you may be able to run your jobs anyway, if a sufficient number of nodes happens to be free on the system.
  • Very wide jobs will only be started within a maintenance window (just before the maintenance window or at the end of the maintenance window). These are planned for the first Wednesday of each month. On Snowy a "very wide" job asks for 100 nodes or more.

Convenience Variables

  • $SNIC_TMP - Path to node-local temporary disk space

    The $SNIC_TMP variable contains the path to a node-local temporary file directory that you can use when running your jobs, in order to get maxiumum disk performance (since the disks are local to the current compute node). This directory will be automatically created on your (first) compute node before the job starts and automatically deleted when the job has finished.

    The path specified in $SNIC_TMP is equal to the path: /scratch/$SLURM_JOB_ID, where the job variable $SLURM_JOB_ID contains the unique job identifier of your job.

    WARNING: Please note, that in your "core" (see below) jobs, if you write data in the /scratch directory but outside of the /scratch/$SLURM_JOB_ID directory, your data may be automatically deleteted during your job run.
 

Details about the "core" and "node" partitions

A normal Snowy node contains 128 GB of RAM and 16 compute cores. An equal share of RAM for each core would mean that each core gets at most 8 GB of RAM. This simple calculation gives one of the limits mentioned below for a "core" job.

You need to choose between running a "core" job or a "node" job. A "core" job must keep within certain limits, to be able to run together with up to 19 other "core" jobs on a shared node. A job that cannot keep within those limits must run as a "node" job.

Some serial jobs must run as "node" jobs. You tell Slurm that you need a "node" job with the flag "-p node". (If you forget to tell Slurm, you are by default choosing to run a "core" job.)

A "core" job:

  • Will use a part of the resources on a node, from a 1/16 share to a 15/16 share of a node.

  • Must specify less cores than 16, i.e.between "-n 1" to "-n 15".

  • Must not demand "-N", "--nodes", or "--exclusive".

  • Is recommended not to demand "--mem"

  • Must not demand to run on a fat node (see below, for an explanation of "fat"), a devel node.

  • Must not use more than 8 GB of RAM for each core it demands. If a job needs half of the RAM, i.e. 64 GB, you need to reserve also at least half of the cores on the node, i.e. 8 cores, with the "-n 8" flag.

A "core" job is accounted on your project as one "core hour" (sometimes also named as a "CPU hour") per core you have been allocated, for each wallclock hour that it runs. On the other hand, a "node" job is accounted on your project as sixteen core hours for each wallclock hour that it runs, multiplied with the number of nodes that you have asked for.

Node types

Rackham has two node types, thin being the typical cluster node and fat nodes having double the amount of memory available normally (256 GB). You may specify a node with more RAM, by adding the words "-C mem256GB" or "-C fat" to your job submission line and thus making sure that you will get 256 GB of RAM on each node in your job. If you need even more memory you can request a 512 GB memory node by adding "-C mem512GB". Please note that there are only 13 nodes with 256GB and 17 with 512GB of RAM.

To request a fat node, use -C mem256GB or -C fat in your Slurm command. 

To request the fattest nodes, use -C mem512GB in your Slurm command. 

File storage and disk space

At UPPMAX we have a few different kinds of storage areas for files, see Disk Storage User Guide for more information and recommended use.

Very Long jobs

If you have very long jobs that require more than 10 days of CPU-time. We recommend using Snowy.
But in order for our job to successfully run for severay weeks you should implement the following;

  • Use only local disk for your job. Copy all input and data files needed to $SNIC_TMP at the start of your job and at the end, copy all output back to your project directory.
  • Book a full node with the SLURM flags -p node -n 16
  • If possible copy the program files needed to $SNIC_TMP

Even if you do this we can't promise that a 20 or 30 day long job will finish without being interuppted by the global file systems, the network or problems on the node,