Practical 10:

Music Track Analytics using MapReduce

Objective:
Compute, per track:

• Number of unique listeners
• Total shares
• Radio plays
• Total listens

Input: music_logs.txt (pipe-separated log file):

UserId | TrackId | Shared | Radio | Skip

111115 | 222 | 0 | 1 | 0

111113 | 225 | 1 | 0 | 0

111117 | 223 | 0 | 1 | 1

111115 | 225 | 1 | 0 | 0

111118 | 222 | 0 | 0 | 0

111119 | 223 | 0 | 1 | 0

111115 | 222 | 0 | 0 | 0

Steps:

1.     Upload Input to HDFS

hdfs dfs -mkdir /music_input

hdfs dfs -put music_logs.txt /music_input/music_logs.txt

2.     Custom Writable Classes

• TrackActivityWritable: holds per-record activity flags and userId
• TrackStatsWritable: holds final aggregated metrics

3.     Mapper:

• Skip header
• Parse each log line
• Emit (TrackId, TrackActivityWritable)

4.     Reducer:

• Aggregate metrics per track:
• Unique listeners (HashSet)
• Total listens
• Shared count
• Radio plays
• Emit (TrackId, TrackStatsWritable)

5.     Driver:

• Set Mapper, Reducer, Output Key/Value classes
• Specify input/output paths
• Submit job

6.     Compile & Package

mkdir -p classes

javac -cp "$(hadoop classpath)" -d classes *.java

jar -cvf music_analytics.jar -C classes/ .

7.     Run Job

hdfs dfs -rm -r /music_output

hadoop jar music_analytics.jar MusicAnalyticsDriver /music_input/music_logs.txt /music_output

8.     Verify Output

hdfs dfs -cat /music_output/part-r-00000

Example Output:

222  2  0  1  3

223  2  0  2  2

225  2  2  0  2

(Columns: TrackId, UniqueListeners, SharedCount, RadioListens, TotalListens)

Conclusion:
Students now know how to:

• Use custom Writable classes for complex data structures
• Aggregate multiple metrics in a single MapReduce job
• Extract meaningful analytics from raw user logs