1. Compute Service Configuration

In order to utilize a SIMA Compute Service to execute workflows from SIMA, the connection must first be defined. This is done in the preferences dialog, under SIMA Compute.

Common for all connections are the notification settings. Here you specify an e-mail address where a message will be sent when a workflow execution has ended. Either because it has completed successfully or because it has failed in some way.

1.1. HPC Configuration

The Maximum number of concurrent condition runs setting determines how many SIMA conditions can run concurrently for each SIMA Runtime Engine instance. This value should normally be left at 1 when running on a cluster.

The Maximum number of concurrent workflow runs setting determines how many SIMA workflow can run concurrently for each SIMA Runtime Engine instance. This value should normally be left at 1 when running on a cluster.

The Maximum array job size setting determines how a workflow set is split up into smaller parts so that it can be computed concurrently. If you have a workflow set with 2000 permutations and array size of 200 you will have 200 jobs on the cluster, each computing 10 permutations. The default value here is equal to the maximum allowed on the Slurm resource manager, 1000.

The Concurrent array job executions determines how many of one specific array job can execute at one time. So if you have 200 jobs in one array and this value is set to 10, only ten of the jobs can execute concurrently regardless of how many free cores there are. The default is to execute as many jobs as possible and leave it up to the cluster resource manager to decide exactly how many.

Allowed memory usage and CPU time per job are used to limit how much memory and CPU time each job requires. If left at auto the job will use whatever allowed by the queue on the cluster. Some installations may have a default value much lower than the maximum allowed value for the queue. In this case one should specify the requirements here.

The HPC queue/partition setting determines which queue to use on the cluster. The default value here is to use whatever is configured at the SIMA compute service installation.