Use Tall Arrays on a Spark Cluster

This example shows how to modify a MATLAB^® example of creating a tall table to run on a Spark™ cluster or a Spark enabled Hadoop^® cluster. You can use this tall table to create tall arrays and calculate statistical properties. You can develop code locally and then scale up, to take advantage of the capabilities offered by Parallel Computing Toolbox™ and MATLAB Parallel Server™ without having to rewrite your algorithm. See also Big Data Workflow Using Tall Arrays and Datastores, Configure a Spark Cluster (MATLAB Parallel Server), and Configure a Hadoop Cluster (MATLAB Parallel Server)

Set Up a Spark Cluster and a Spark Enabled Hadoop Cluster

First, you must set environment variables and cluster properties as appropriate for your specific Spark cluster configuration. See your system administrator for the values for these and other properties necessary for submitting jobs to your cluster.

Set Cluster Properties on a Spark Cluster

To set cluster properties on Spark cluster, type the following commands in the MATLAB command window:

cluster = parallel.cluster.Spark('SparkInstallFolder', '/path/to/spark/install');

% Optionally, if you want to control the exact number of workers:
cluster.SparkProperties('spark.executor.instances') = '16';

mapreducer(cluster);

Set environment variables and cluster properties on a Spark enabled Hadoop Cluster

To set environment variable and cluster properties on a Spark enabled Hadoop Cluster, type the following commands in the MATLAB command window:

setenv('HADOOP_HOME', '/path/to/hadoop/install')
setenv('SPARK_HOME', '/path/to/spark/install');
cluster = parallel.cluster.Hadoop;

% Optionally, if you want to control the exact number of workers:
cluster.SparkProperties('spark.executor.instances') = '16';

mapreducer(cluster);

Note

In the setup step, you use mapreducer to set the cluster execution environment. In the next step, you create a tall array. If you modify or delete the cluster execution environment after creating a tall array, then the tall array is invalid and you must recreate it.

Note

If you want to develop in serial and not use local workers, enter the following command.

mapreducer(0);

Creating and Using Tall Tables

After setting your environment variables and cluster properties, you can run the MATLAB tall table example on your Spark cluster instead of on your local machine.

These instructions show how to create and use tall tables on a Spark enabled Hadoop cluster, although this procedure can be used for any Spark cluster.

Create a datastore and convert it into a tall table. MATLAB automatically starts a Spark job to run subsequent calculations on the tall table.

ds = datastore('airlinesmall.csv');
varnames = {'ArrDelay', 'DepDelay'};
ds.SelectedVariableNames = varnames;
ds.TreatAsMissing = 'NA';

Create a tall table tt from the datastore.

tt = tall(ds)

Starting a Spark Job on the Hadoop cluster. This could take a few minutes ...done.

tt =

  M×2 tall table 

    ArrDelay    DepDelay
    ________    ________

     8          12      
     8           1      
    21          20      
    13          12      
     4          -1      
    59          63      
     3          -2      
    11          -1      
    :           :
    :           :

The display indicates that the number of rows, M, is not yet known. M is a placeholder until the calculation completes.

Extract the arrival delay ArrDelay from the tall table. This action creates a new tall array variable to use in subsequent calculations.

a = tt.ArrDelay;

You can specify a series of operations on your tall array, which are not executed until you call gather. Doing so allows you to batch up commands that might take a long time. As an example, calculate the mean and standard deviation of the arrival delay. Use these values to construct the upper and lower thresholds for delays that are within 1 standard deviation of the mean.

m = mean(a,'omitnan');
s = std(a,'omitnan');
one_sigma_bounds = [m-s m m+s];

Use gather to calculate one_sigma_bounds, and bring the answer into memory.

sig1 = gather(one_sigma_bounds)

Evaluating tall expression using the Spark Cluster:
- Pass 1 of 1: Completed in 0.95 sec
Evaluation completed in 1.3 sec

sig1 =

  -23.4572    7.1201   37.6975

You can specify multiple inputs and outputs to gather if you want to evaluate several things at once. Doing so is faster than calling gather separately on each tall array. For example, calculate the minimum and maximum arrival delay.

[max_delay, min_delay] = gather(max(a),min(a))

max_delay =

        1014

min_delay =

   -64

Note

These examples take more time to complete the first time if MATLAB is starting on the cluster workers.

When using tall arrays on a Spark cluster, compute resources from the cluster will be reserved for the lifetime of the mapreducer execution environment. To clear these resources, you must delete the mapreducer:

delete(gcmr);

Alternatively, you can change to a different execution environment, for example:

mapreducer(0);