If you do not have an existing scheduler in your cluster, follow these instructions to integrate the MATLAB® Job Scheduler, which is provided with MATLAB Parallel Server™. If you already have a cluster with a scheduler, see Integrate MATLAB with Third-Party Schedulers.
After you integrate MATLAB with a scheduler, you can access workers in your cluster from a desktop MATLAB client session (requires Parallel Computing Toolbox™).
The setup in these steps uses online licensing.
To install MATLAB Parallel Server using online licensing, you must check your license type.
In your browser, go to License Center and log in with your MathWorks® Administrator Account.
Select the MATLAB Parallel Server license that you plan to use.
On the Install and Activate tab, look for License Manager: followed by the license manager type currently assigned to this license.
Next, define the end users permitted to check out the license.
If you are not already logged in as an administrator, go to License Center and log in with your MathWorks Administrator Account.
Select your MATLAB Parallel Server license, and then click Manage Users.
Click Add User to add a user to the list.
Provide the user’s email address, first and last names, and country. Click Add User. Note that if the specified email address does not correspond to an existing MathWorks Account, a new account is created for that user.
Add end users as necessary.
To save time and eliminate the need to run the installer-based download process on each computer in your cluster, download the installation files prior to installation. Doing so facilitates installation in a large number of machines. If you have access to an Administrator’s account for your license, you can use the installer to download files without installing them. If not, contact the administrator of your license to obtain a copy of the installation files. For more information, see Download Products Without Installing (Installation and Licensing). When you download the files using the installer, you must:
Select the operating system for the cluster machines.
Select all products for download. MATLAB Parallel Server cannot run jobs requiring products that are not installed.
MATLAB Parallel Server has two server-side components:
To install the software on each computer in your cluster, follow these steps:
Start the MATLAB installer from the installation files downloaded in Get the Installation Files.
Select Log in with a MathWorks Account and follow the prompts.
Select all products that the end users will use, and exclude the license manager.
After the installation completes, update the
matlabroot/toolbox/parallel/bin. Uncomment and
For best performance, install the software locally on each node. However, you can also install the software in a network share location. For your convenience, you can perform noninteractive installation (silent installation) on the worker nodes. For instructions, see Install Noninteractively (Installation and Licensing).
The MATLAB Job Scheduler is a scheduler that is provided with MATLAB Parallel Server. The MATLAB Job Scheduler is intended primarily for small-to-medium-sized clusters that run only MATLAB jobs. The scheduler interface is a high-level abstraction that enables you to submit jobs to your computation resources, and allows you to avoid dealing with differences in operating systems and environments.
On the head node, start Admin Center. Go to
matlabroot/toolbox/parallel/bin and execute the file
matlabroot is the
MATLAB installation folder.
Click Add or Find, and specify the computers that you are using as your head node and worker nodes.
Follow the prompts and confirm to start the mjs service. If necessary, manually start the mjs service using the command-line interface. For more information, see Use the Command-Line Interface (Windows) or Use the Command-Line Interface (UNIX).
In the MATLAB Job Scheduler section, click Start. Specify a name for your MATLAB Job Scheduler and select the head node from the dropdown list.
To add MATLAB Parallel Server workers, click Start in the Workers section of the Admin Center.
Select the computers to host the workers.
Select the number of workers per computer.
Verify your configuration by checking worker status in the Workers section.
To troubleshoot issues, click Test Connectivity in the Host section.
If you use UNIX®, configure the mjs service to start automatically at start time. For instructions, see Start the mjs Service, MATLAB Job Scheduler, and Workers (Command-Line).
The following screenshot shows the final setup in Admin Center.
If you need more help during the configuration, such as your cluster requires firewall configuration or you want to set up multiple mjs installations, see this more detailed guide Configure Advanced Options for MATLAB Job Scheduler Integration.
To use MATLAB Parallel Server, you must have a client computer running MATLAB and Parallel Computing Toolbox. In the MATLAB toolstrip, select Parallel > Discover Clusters and follow the instructions to automatically discover and set up your cluster.
Alternatively, you can configure it manually as follows:
In MATLAB, on the Home tab, select Parallel > Create and Manage Clusters.
Select Add Cluster Profile > MATLAB Job Scheduler.
Create your MATLAB Job Scheduler profile and click Edit.
Update the hostname of the head node.
Update the license number.
Click Done and select Set as Default (optional) .
After you successfully validate your cluster, you can now use your MATLAB session to submit jobs to the MATLAB Parallel Server cluster.
If your validation does not pass, contact the MathWorks install support team.
Any MATLAB cluster workers that you start use dynamic licensing: they can use all the functionality you are licensed for in the MATLAB client, while checking out only MATLAB Parallel Server licenses in the cluster.
For information on configuring more advanced options for your cluster, see MATLAB Job Scheduler Cluster Customization. For example, you can set the security of the cluster in Set MATLAB Job Scheduler Cluster Security. After you finish your configuration, try some examples of cluster workflows in Running Code on Clusters and Clouds.