Abaqus

Abaqus FEA is a software suite for finite element analysis and computer-aided engineering.

Using your own license

Abaqus is available on Compute Canada clusters, but you must provide your own license. To configure your cluster account, create a file named $HOME/.licenses/abaqus.lic corresponding to the module version you want to use (otherwise abaqus jobs will fail). This must be done on each cluster where you plan to run abaqus as follows:

Module Version 2020

File : abaqus.lic

prepend_path("ABAQUSLM_LICENSE_FILE","port@server")

Module Version 6.14.1

File : abaqus.lic

prepend_path("LM_LICENSE_FILE","port@server")

Replace port@server with the port number and name of your Abaqus license server. Your license server must be reachable by our compute nodes, so your firewall will need to be configured appropriately. This usually requires our technical team to get in touch with the technical people managing your license software. Please contact our technical support and we will provide a list of IP addresses used by our clusters and obtain the information we need on the port and IP address of your server.

Cluster job submission

Below is a sample slurm script to submit a parallel job to a single compute node using 4 cores:

File : "abaqus_job.sh"

#!/bin/bash
#SBATCH --time=00-06:00        # days-hrs:mins
#SBATCH --mem=8G               # node memory > 5G
#SBATCH --cpus-per-task=4      # number cores > 1
 
module load abaqus/6.14.1      # (or abaqus/2020)
unset SLURM_GTIDS
export MPI_IC_ORDER='tcp'

abaqus job=test input=sample.inp scratch=$SCRATCH cpus=$SLURM_CPUS_ON_NODE \
   interactive mp_mode=threads memory="$((${SLURM_MEM_PER_NODE}-3072))MB"

where a listing of abaqus options can be obtained by loading an abaqus module and running: abaqus -help | less

Node memory

An estimate for the total slurm node memory (--mem=) required for a simulation to run fully in ram (without being virtualized to scratch disk) can be obtained by examining the abaqus output test.dat file. For example a simulation that requires a fairly large amount of memory might show:

                   M E M O R Y   E S T I M A T E
  
 PROCESS      FLOATING PT       MINIMUM MEMORY        MEMORY TO
              OPERATIONS           REQUIRED          MINIMIZE I/O
             PER ITERATION           (MB)               (MB)
  
     1          1.89E+14             3612              96345

To run your simulation interactively and monitor the memory consumption do the following:

 1) ssh into a compute canada cluster, obtain an allocation on a compute node (such as gra100), run abaqus ie)
    salloc --time=0:30:00 --cpus-per-task=8 --mem=64G --account=def-piname
    module load abaqus/6.14.1  OR  module load abaqus/2020
    unset SLURM_GTIDS
    abaqus job=test input=Sample.inp scratch=$SCRATCH cpus=8 mp_mode=threads interactive
 2) ssh into the compute canada cluster again, ssh into the compute node with the allocation, run top ie)
    ssh gra100
    top -u $USER
 3) watch the VITR and RES columns until steady memory values are observed

To completely satisfy the recommended "MEMORY TO OPERATIONS REQUIRED MINIMIZE I/O" (MRMIO) value at least the same smount of non-swapped physical memory (RES) must be available to abaqus. Since the RES will in general be less than the virtual memory (VIRT) by some relatively constant amount for a given simulation, it is necessary to slightly over allocate the requested slurm node memory -mem=. In the above sample slurm script this over-allocation has been hardcoded to a conservative value of 3072MB based on initial testing of the standard abaqus solver. To avoid long queue wait times associated with large values of MRMIO, it maybe worth investigating the simulation performance impact associated with reducing the RES memory that is made available to abaqus significantly below the MRMIO. This can be done by lowering the -mem= value which in turn will set an artificially low value of memory= in the abaqus command (found in the last line of the slurm script). In doing this one should be careful the RES does not dip below the "MINIMUM MEMORY REQUIRED" (MMR) otherwise abaqus will exit due to "Out Of Memory" (OOM). As an example, if your MRMIO is 96GB try running a series of short test jobs with #SBATCH --mem=8G, 16G, 32G, 64G until an acceptable minimal performance impact is found, noting that smaller values will result in increasingly larger scratch space use by tmpdir files.

Cluster graphical use

Abaqus/2020 can be run interactively in graphical mode on a cluster compute node (3hr time limit) over TigerVNC with these steps:

1. Install TigerVNC client on your desktop
2. Connect to a cluster compute node with vncviewer
3. module load abaqus/2020
4. abaqus cae -mesa

Gra-vdi graphical use

NOTE: gra-vdi is currently OFFLINE for upgrading with a return to use date sometime in june

Abaqus/2020 can be run interactively in graphical mode on gra-vdi (no connection time limit) over TigerVNC with these steps:

1. Install TigerVNC client on your desktop
2. Connect to gra-vdi.computecanada.ca with vncviewer
3. module load SnEnv
4. module load abaqus/2020
5. abaqus cae

Site specific usage

Sharcnet license

Sharcnet provides a small but free license consisting of 2cae and 21 execute tokens where usage limits are imposed 10 tokens/user and 15 tokens/group. For groups that have purchased tokens, the free token usage limits are added to their reservation. The free tokens are available on a first come first serve basis and mainly intended for testing and light usage before deciding whether or not to purchase dedicated tokens. The license can be used by any Compute Canada member but only on SHARCNET hardware. Groups that purchase dedicated tokens to run on the SHARCNET license server may likewise only use them on SHARCNET hardware. Such hardware includes gra-vdi for running abaqus in full graphical mode and graham cluster for submitting compute batch jobs to the queue. Before you can use the license you must open ticket at <support@computecanada.ca> and request access. In your email 1) mention that it is for use on Sharcnet systems and 2) include a copy/paste of the following License Agreement statement with your full name and Compute Canada username entered in the indicated locations. Please note that every user must do this ie) cannot be done one time only for a group (including PIs who have purchased their own dedicated tokens).

o License agreement

----------------------------------------------------------------------------------
Subject: Abaqus Sharcnet Academic License User Agreement

This email is to confirm that i "_____________" with username "___________" will
only use “SIMULIA Academic Software” with tokens from the SHARCNET license server
for the following purposes:

1) on SHARCNET hardware where the software is already installed
2) in affiliation with a canadian degree-granting academic institution
3) for education, institutional or instruction purposes and not for any commercial
   or contract related purposes where results are not publishable
4) for experimental, theoretical and/or digital research work, undertaken primarily
   to acquire new knowledge of the underlying foundations of phenomena and observable
   facts, up to the point of proof-of-concept in a laboratory    
-----------------------------------------------------------------------------------

o Configure license file

The configuration of your abaqus license on each cluster depends on the module version being used:

Module Version 2020

[gra-login1:~] cat ~/.licenses/abaqus.lic
prepend_path("ABAQUSLM_LICENSE_FILE","27050@license3.sharcnet.ca")

Module Version 6.14.1

[gra-login1:~] cat ~/.licenses/abaqus.lic
prepend_path("LM_LICENSE_FILE","27050@license3.sharcnet.ca")

If your abaqus jobs fail with error message [*** ABAQUS/eliT_CheckLicense rank 0 terminated by signal 11 (Segmentation fault)] in the slurm output file verify your abaqus.lic file contains ABAQUSLM_LICENSE_FILE to use abaqus/2020. If your abaqus jobs fail with error message starting [License server machine is down or not responding etc] in the output file verify your abaqus.lic file contains LM_LICENSE_FILE to use abaqus/6.14.1 as shown.

o Check license status

To query the sharcnet license server for started jobs, queued jobs, and reservations by purchasing groups run:

ssh graham.computecanada.ca
module load abaqus
abaqus licensing lmstat -c $LM_LICENSE_FILE -a | grep "Users\|start\|queued\|RESERVATIONs"

When abaqus licensing lmstat shows your job is "queued" this means it has entered the "R"unning state from the perspective of either the squeue -j jobid or sacct -j jobid commands but is waiting for a license (not started) and therefore idle. This will have the same impact on your account priority as if the job were consuming cputime and thus should be avoided. When abaqus licensing lmstat indicates the job is in the "start" state then it has acquired the required abaqus tokens from the license and consuming cputime ...

o Specify job resources

To ensure optimal usage of both your Abaqus tokens and the Compute Canada resources its important to carefully specify the required memory and ncpus in your slurm script. The values can be determined by submitting a few short test jobs to the queue then checking their utilization. For completed jobs use seff JobNumber to show the total "Memory Utilized" and "Memory Efficiency"; If the "Memory Efficiency" is less than ~90% decrease the value of "#SBATCH --mem=" setting in your slurm script accordingly. Notice that the seff JobNumber command also shows the total "CPU (time) Utilized" and "CPU Efficiency"; If the "CPU Efficiency" is less than ~90% perform scaling tests to determine the optimal number of cpu's for optimal performance and then update the value of then update the value of "#SBATCH --cpus-per-task=" in your slurm script. For running jobs use the srun --jobid=29821580 --pty top -d 5 -u $USER command to watch the %CPU, %MEM and RES for each abaqus parent process on the compute node; The %CPU and %MEM columns display the percent usage relative to the total available on the node while the RES column shows the per process resident memory size (in human readable format for values over 1gb). Further information regarding howto Monitor Jobs is available in the Compute Canada wiki.

o Core token mapping

TOKENS 5  6  7  8  10  12  14  16  19  21  25  28  34  38
CORES  1  2  3  4   6   8  12  16  24  32  48  64  96 128

where TOKENS = floor[5 X CORES^0.422]

o Using your license

To use your own license server (instead of the default SHARCNET license) on gra-vdi as described in the "Gra-vdi graphical use" section above, run command export ABAQUSLM_LICENSE_FILE="port@server" after loading the abaqus module and before running abaqus cae.

Western license

The Western site license may only be used by Western researchers with hardware located on Western's campus such as the Dusky legacy cluster. Graham and gra-vdi are excluded since they are located at Waterloo (use the Sharcnet License for these systems as described above). Contact the Western abaqus license server administrator (located in Robarts) to make arrangements before attempting to use the Western abaqus license. Submit a ticket to Compute Canada support to request the admins contact information if necessary. You will need to provide your Compute Canada username and likely make arrangements to purchase tokens. If you are granted access request the port and server values and enter them into your abaqus.lic file as shown in 1) near the top of this wiki which will in turn be used by the Compute Canada module on dusky when it loads.

Message Passing Interface