java - MapReduce to Spark

Question

I have a MapReduce job written in Java. It depends on multiple classes. I want to run the MapReduce job on Spark.

What steps should I follow to do the same?

I need to make changes only to the MapReduce class?

Thanks!

Answer 1

This is a very broad question, but the short of it is:

1. Create an RDD of the input data.
2. Call map with your mapper code. Output key-value pairs.
3. Call reduceByKey with your reducer code.
4. Write the resulting RDD to disk.

Spark is more flexible than MapReduce: there is a great variety of methods that you could use between steps 1 and 4 to transform the data.