This topic describes how to use MapReduce to read and write data in JindoFileSystem (JindoFS).

JindoFS configuration

For example, a namespace named emr-jfs is created with the following configuration:

  • jfs.namespaces=emr-jfs
  • jfs.namespaces.emr-jfs.uri=oss://oss-bucket/oss-dir
  • jfs.namespaces.emr-jfs.mode=block

MapReduce introduction

Generally, Hadoop MapReduce jobs read and write data through Hadoop Distributed File System (HDFS). JindoFS is compatible with HDFS. You can set the input and output directories of MapReduce jobs to directories in JindoFS. In this case, MapReduce jobs can read and write data in JindoFS.

Hadoop MapReduce is a software framework for easily writing applications. Applications based on Hadoop MapReduce can process multi-terabyte datasets in parallel in large clusters that consist of thousands of nodes in a reliable and fault-tolerant manner. A MapReduce job usually splits the input dataset into independent blocks that are processed by map tasks in parallel. The MapReduce framework sorts the output of map tasks and writes the sorted output to reduce tasks. Both the input and the output of the job are stored in a file system. The MapReduce framework is responsible for scheduling tasks, monitoring tasks, and rerunning failed tasks.

Job input and output

In applications, the input and output directories of MapReduce jobs are specified. Map and reduce functions are implemented based on appropriate methods or abstract classes. The Hadoop job client submits MapReduce jobs and their configuration to ResourceManager. Then, ResourceManager schedules tasks. In this case, you can modify the input and output directories of MapReduce jobs to directories in JindoFS to read and write data in JindoFS.

Examples

The following examples show how to modify the input and output directories of a MapReduce job to read and write data in JindoFS.

  • Teragen

    In this MapReduce job, use Teragen to generate data in a specified number of rows in the specified directory:

    hadoop jar /usr/lib/hadoop-current/share/hadoop/mapreduce/hadoop-mapreduce-examples-*.jar teragen <num rows> <output dir>

    Replace the output directory with a directory in JindoFS to write data to JindoFS:

    hadoop jar /usr/lib/hadoop-current/share/hadoop/mapreduce/hadoop-mapreduce-examples-*.jar teragen 100000 jfs://emr-jfs/teragen_data_0
  • Terasort

    In this MapReduce job, use Terasort to sort data in the specified input directory and write the sorted data to the specified output directory:

    hadoop jar /usr/lib/hadoop-current/share/hadoop/mapreduce/hadoop-mapreduce-examples-*.jar terasort <in> <out>

    Replace the input and output directories with directories in JindoFS to process data in JindoFS:

    hadoop jar /usr/lib/hadoop-current/share/hadoop/mapreduce/hadoop-mapreduce-examples-*.jar terasort jfs://emr-jfs/teragen_data_0/  jfs://emr-jfs/terasort_data_0