Working with MapReduce jobs

MapReduce is a parallel processing tool for large datasets (a few dozen TB or more) on clusters in the Hadoop ecosystem. It allows you to work with data in different formats. Job input and output are stored in Yandex Object Storage. MapReduce uses a number of libraries, while Apache Bigtop determines the path to them.

In this article, we use a simple example to show how MapReduce works in Yandex Data Processing. We will use MapReduce to compute the population of the world's 500 largest cities based on a set of data on the cities.

To run MapReduce on Hadoop, we use the Streaming interface. At the same time, the data preprocessing (map) and the final output computation (reduce) stages use programs that read data from a standard program input (stdin) and write their output to a standard output (stdout).

Getting startedGetting started

1. Create a service account with the dataproc.agent and dataproc.provisioner roles.

2. In Object Storage, create buckets and configure access to them:

   1. Create a bucket for the input data and grant the READ permission for this bucket to the cluster service account.
   2. Create a bucket for the processing output and grant the cluster service account READ and WRITE permissions for this bucket.

3. Create a Yandex Data Processing cluster with the following settings:

   • Environment: PRODUCTION
   • Services:
     • HDFS
     • MAPREDUCE
     • YARN
   • Service account: Select the service account you previously created.
   • Bucket name: Select a bucket to hold the processing results.

Create a MapReduce jobCreate a MapReduce job

1. Download an archived CSV file with a dataset on the cities and upload it to the input data bucket.

2. Upload Python files to the input data bucket: mapper.py, which contains the code for data preprocessing (map stage), and reducer.py, which contains the code for the final computations (reduce stage):

   mapper.py

   #!/usr/bin/python
   import sys
   
   population = sum(int(line.split('\t')[14]) for line in sys.stdin)
   print(population)
   

   reducer.py

   #!/usr/bin/python
   import sys
   
   population = sum(int(value) for value in sys.stdin)
   print(population)
   

3. Create a MapReduce job with the following parameters:

   • Main class: org.apache.hadoop.streaming.HadoopStreaming
   • Arguments:
     • -mapper
     • mapper.py
     • -reducer
     • reducer.py
     • -numReduceTasks
     • 1
     • -input
     • s3a://<input_data_bucket_name>/cities500.txt
     • -output
     • s3a://<output_bucket_name>/<output_directory>
   • Files:
     • s3a://<input_data_bucket_name>/mapper.py
     • s3a://<input_data_bucket_name>/reducer.py
   • Properties:
     • mapreduce.job.maps: 6
     • yarn.app.mapreduce.am.resource.mb: 2048
     • yarn.app.mapreduce.am.command-opts: -Xmx2048m

4. Wait for the job status to change to Done.

5. Download from the bucket and review the file with the result from the bucket:

   part-00000

   3157107417
   

Delete the resources you createdDelete the resources you created

Some resources are not free of charge. To avoid paying for them, delete the resources you no longer need:

1. Delete the cluster.
2. Delete the buckets.
3. Delete the service account.