Configure for Spark Clusters

Submit parallel MATLAB® code that contains tall (MATLAB) arrays and mapreduce (MATLAB) functions to a Spark™ cluster from suitably configured MATLAB clients.

Cluster Configuration

Follow these instruction to interface MATLAB Parallel Server™ installation into your Spark cluster.

  1. Integrate MATLAB Parallel Server with your cluster infrastructure. For instructions, see Install MATLAB Parallel Server for Other Third-Party Schedulers.

    MATLAB Parallel Server supports Spark clusters running in different environments such as Spark Standalone or Databricks™.

  2. If your cluster requires Kerberos authentication, ensure your MATLAB Parallel Server installation have been configured correctly. For instructions, see Kerberos Authentication.

Client Configuration

To configure the client to run MATLAB code on the cluster, you must already be able to submit to the cluster from the intended client machine. The client machine must have a Spark installation that can access the cluster outside of MATLAB.

Many Spark distributions do not support direct access of Linux® based clusters from Windows® clients. Users of Windows clients typically need to set up a Linux gateway node that can be accessed from the Windows client via SSH or VNC. The cluster can then be accessed from this gateway node.

You must ensure your MATLAB client installation has been configured for Kerberos authentication if your cluster requires it. For instructions, see Kerberos Authentication.

Create a Cluster Profile

In this step you create a cluster profile that enables the MATLAB client to connect to the cluster.

  1. Start the Cluster Profile Manager. On the Home tab, in the Environment area, select Parallel > Create and Manage Clusters.

  2. Create a new profile in the Cluster Profile Manager by selecting Add Cluster Profile > Spark.

  3. With the new profile selected in the list, click Rename and edit the profile name to be SparkInstallTest. Press Enter.

  4. In the Properties tab, select Edit and provide settings for the following fields:

    1. Set the Description field to For testing installation.

    2. Set ClusterMatlabRoot to the installation location of MATLAB to run on the worker machines.

    3. If the cluster uses online licensing, set RequiresOnlineLicensing to true.

    4. If you set RequiresOnlineLicensing to true, in the LicenseNumber field, enter your license number.

    5. Set SparkInstallFolder to the Spark installation location on the client machine.

    6. Specify additional Spark properties in the SparkProperties table. For example, to specify the preferred number of workers to 16, under the SparkProperties table, select Add. Specify a new property with the name spark.executor.instances, enter the 16 for the value, and select String as the data type.

    7. Click Done to save your cluster profile changes. The dialog box looks as follows:

      Cluster Profile Manager with the SparkInstallTest cluster profile selected. The properties of the InstallTest profile are displayed.

Validate the Cluster Profile

In this step you verify your cluster profile, and thereby your installation.

  1. If it is not already open, start the Cluster Profile Manager from the MATLAB desktop. On the Home tab, in the Environment area, select Parallel > Create and Manage Clusters.

  2. Select the SparkInstallTest cluster profile in the listing.

  3. Click Validation tab.

  4. Use the checkboxes to choose all tests, or a subset of the validation stages, and specify the number of workers to use when validating your profile.

  5. Click Validate.

After the client completes the cluster validation, the Validation tab shows the output. The following figure shows the results of a profile that passed all validation tests.

Cluster Profile Manager with the SparkInstallTest profile selected. The validation results for the Spark cluster are shown in the right pane.

Note

If your validation does not pass, contact the MathWorks install support team.

If your validation passed, you now have a valid profile that you can use to access the cluster from a MATLAB client. You can make any modifications to your profile appropriate for your applications, such as specifying additional Spark properties, NumWorkersRange, AttachedFiles, AdditionalPaths, and so on.

To save your profile for other users, select the profile and click Export, then save your profile to a file in a convenient location. Later, when running the Cluster Profile Manager, other users can import your profile by clicking Import.

Examples

Use mapreducer (MATLAB) function to change the execution environment to the Spark cluster.

For examples of how to run parallel MATLAB code on your Spark cluster, see Use Tall Arrays on a Spark Cluster (Parallel Computing Toolbox).

Kerberos Authentication

If the cluster uses Kerberos authentication that requires the Oracle® Java® Cryptography Extension, you must configure all installations of MATLAB and MATLAB Parallel Server. If you are using Spark on Hadoop®, for example Cloudera® distributions, it is likely that you need to complete these configuration steps.

The configuration instructions are the same for client and worker MATLAB installations.

Starting in R2018b, configure your MATLAB installation by enabling the appropriate security policy in the Java installation.

  1. In the MATLAB Editor, open the file ${MATLAB_ROOT}/sys/java/jre/${ARCH}/jre/lib/security/java.security.

  2. Change the line 

    #crypto.policy=unlimited
    to
    crypto.policy=unlimited

Spark Version Support

Spark 2.2 or later supports MATLAB mapreduce, tall arrays and parallel usage of datastores. For the client, you can use tall arrays on Spark clusters supporting all architectures, while supporting Linux and Mac architectures for the cluster. This includes cross-platform support.

Spark Executor Memory Default

The default value of the spark.executor.memory property of a Spark job submitted from MATLAB is 2560 MB.

See Also

(Parallel Computing Toolbox)

Topics