4.2. Overview of an Ansys RSM Cluster (ARC) Configuration

An Ansys RSM Cluster (ARC) can be a single- or multi-node cluster.

In a single-node ARC, the head node (submit host) and execution node are the same machine. This can be a user's local machine (see The Default 'Localhost' Configuration), or a single remote machine to which multiple users submit jobs.

A multi-node ARC consists of two or more machines. It is based on the head/compute model of communication in which one machine or process has control over the other machines. In this scenario one machine serves as the head node (submit host), and the other machines serve as execution nodes. The submit host may also serve as an execution node if desired.

In order for an ARC to be operational (that is, able to accept, schedule and execute jobs), ARC services must be running on the cluster nodes.


Important:  When a user submits a job to a single-node ARC, RSM will check to see if ARC services are running on that node. If the services are not running, RSM will automatically start them as the user who submitted the job. This is fine for a 'localhost' configuration where a user will be submitting jobs locally to his or her own machine. However, if multiple users will be submitting jobs to an ARC, whether it be a single- or multi-node cluster, we recommend that you install and start ARC services as daemons before anyone submits a job to it.


The steps that you need to take to set up an ARC depend on whether the cluster will be a single- or multi-node cluster, and who will be submitting jobs to it. The table below provides an overview of ARC setup types and the tasks you need to perform for each setup.

You can perform most of these tasks using the automated ARC Configuration application or arcdeploy command. See Creating and Deploying an Ansys RSM Cluster (ARC).

Table 4.1: Overview of ARC Setup Options

ARC TypeARC Service InstallationAdditional Tasks
Single-Node Cluster: Single-User Access (Local Job Submission)
  • User can submit jobs locally to his or her own machine using Local queue

  • Submit host = localhost

RSM will automatically start ARC services on the single node (if they have not already been started)

Single-Node Cluster: Multi-User Access (Remote Job Submission)
  • Multiple users can submit jobs to a specific machine on the network

  • Submit host = machineName.domain.com

  • Jobs execute on the submit host

Before any user submits a job, you must install the ARC Master service and ARC Node service on the single node.

You can start ARC services when using the automated ARC Configuration application or the arcdeploy command. See Creating and Deploying an Ansys RSM Cluster (ARC).

Or, to start ARC services manually, refer to one of the following:

Installing ARC Cluster Services on Windows (installservice)

Installing ARC Cluster Services on Linux

Multi-Node Cluster: Multi-User Access (Remote Job Submission)
  • Multiple users can submit jobs to a specific machine on the network

  • Submit host = machineName.domain.com

  • Two built-in cluster queues: default (includes all machines on which the Node service is installed) and local (contains only the submit host)

  • Submit host associated with one or more execution nodes

  • Jobs execute on the execution node(s) but may also execute on the submit host

Before any user submits a job, you must install the ARC Master service on the submit host (making it the master node), and the ARC Node service on each execution node.

Note:  If the head node will also be used to run jobs, then you must install both the Master service and Node service on that node.

You can start ARC services when using the automated ARC Configuration application or the arcdeploy command. See Creating and Deploying an Ansys RSM Cluster (ARC).

Or, to start ARC services manually, refer to one of the following:

Installing ARC Cluster Services on Windows (installservice)

Installing ARC Cluster Services on Linux

Required

Optional



Important:
  • If the Master service or Node service does not start, consult the ArcMaster242-<date>.log or ArcNode242-<date>.log file to find the possible cause. For more information, see Accessing RSM Log Files.

  • If users will be running a large number of jobs (such as Workbench design point updates) on a single-node Ansys RSM Cluster (ARC), the single node on which jobs are executed could become overloaded, resulting in system issues such as memory usage errors. In this case, it would be advisable to use a multi-node ARC or commercial third-party cluster instead.

  • If you have a firewall set up, this may prevent communication between the master node and execution nodes. To resolve this issue, see Dealing with a Firewall in a Multi-Node Ansys RSM Cluster (ARC).