Reduce phase explained in Hadoop

Introduction

The Reduce phase collects and combines data from the Map phase to produce final results. It helps summarize or aggregate data for easier understanding.

When counting total sales per product from many sales records.

When summing up votes for candidates in an election.

When grouping and averaging temperatures from weather sensors.

When merging logs from multiple servers to find total errors.

Syntax

Hadoop

reduce(key, list_of_values) {
  // process values
  // emit(key, combined_value)
}

The reduce function takes a key and all values for that key from the Map phase.

It combines these values to produce a smaller set of results.

Examples

This example sums all counts for the key 'apple'.

Hadoop

reduce('apple', [2, 3, 5]) {
  sum = 0
  for value in list_of_values:
    sum += value
  emit('apple', sum)
}

This example counts total errors by summing all 1s.

Hadoop

reduce('error', [1, 1, 1, 1]) {
  total_errors = sum(list_of_values)
  emit('error', total_errors)
}

Sample Program

This code simulates the Reduce phase by summing values for each key and printing the result.

Hadoop

def reduce(key, values):
    total = sum(values)
    print(f"{key}: {total}")

# Example data from Map phase
mapped_data = {
    'apple': [2, 3, 5],
    'banana': [1, 1],
    'orange': [4]
}

for key, values in mapped_data.items():
    reduce(key, values)

Important Notes

The Reduce phase only sees data grouped by key from the Map phase.

It is important to write reduce logic that correctly combines all values.

Summary

The Reduce phase combines data from Map outputs by key.

It helps summarize or aggregate large data sets.

Reduce functions take a key and list of values, then emit a combined result.

Practice

1.

What is the main purpose of the Reduce phase in Hadoop MapReduce?

easy

A. To combine and summarize data from the Map outputs by key

B. To split the input data into smaller chunks

C. To read data from the file system

D. To sort the input data before mapping

Solution

1. Step 1: Understand the role of Reduce phase

   The Reduce phase takes the output from the Map phase and groups data by keys.

2. Step 2: Identify the main function of Reduce

   It combines or aggregates the values for each key to produce summarized results.

3. Final Answer:

   To combine and summarize data from the Map outputs by key -> Option A

4. Quick Check:

   Reduce phase = combine by key [OK]

Hint: Reduce phase aggregates data by keys from Map outputs [OK]

Common Mistakes:

• Confusing Reduce with Map phase
• Thinking Reduce splits data
• Assuming Reduce reads input files

2.

Which of the following is the correct signature of a Reduce function in Hadoop MapReduce?

public void reduce(KEY key, Iterable<VALUE> values, Context context) throws IOException, InterruptedException

easy

A. public void reduce(KEY key, VALUE value, Context context) throws IOException, InterruptedException

B. public void map(KEY key, VALUE value, Context context) throws IOException, InterruptedException

C. public void reduce(KEY key, Iterable<VALUE> values, Context context) throws IOException, InterruptedException

D. public void map(KEY key, Iterable<VALUE> values, Context context) throws IOException, InterruptedException

Solution

1. Step 1: Recall Reduce function signature

   The Reduce function receives a key and an iterable list of values for that key.

2. Step 2: Match the correct method signature

   public void reduce(KEY key, Iterable<VALUE> values, Context context) throws IOException, InterruptedException correctly shows the reduce method with key, Iterable values, and context parameters.

3. Final Answer:

   public void reduce(KEY key, Iterable<VALUE> values, Context context) throws IOException, InterruptedException -> Option C

4. Quick Check:

   Reduce signature = key + Iterable values + context [OK]

Hint: Reduce gets key and list of values, not single value [OK]

Common Mistakes:

• Confusing map and reduce method signatures
• Using single VALUE instead of Iterable<VALUE> in reduce
• Mixing map parameters with reduce function

3.

Given the following Reduce function snippet, what will be the output for key = "apple" and values = [2, 3, 5]?

public void reduce(Text key, Iterable<IntWritable> values, Context context) throws IOException, InterruptedException {
    int sum = 0;
    for (IntWritable val : values) {
        sum += val.get();
    }
    context.write(key, new IntWritable(sum));
}

medium

A. (apple, 2)

B. (apple, 10)

C. (apple, 0)

D. (apple, 5)

Solution

1. Step 1: Sum the values list

   Values are 2, 3, and 5. Summing gives 2 + 3 + 5 = 10.

2. Step 2: Output key with summed value

   The context writes the key "apple" with the sum 10 as IntWritable.

3. Final Answer:

   (apple, 10) -> Option B

4. Quick Check:

   Sum of values = 10 [OK]

Hint: Add all values for key to get output [OK]

Common Mistakes:

• Returning only first value instead of sum
• Returning count instead of sum
• Ignoring values and outputting zero

4.

Identify the error in this Reduce function code snippet:

public void reduce(Text key, Iterable<IntWritable> values, Context context) throws IOException, InterruptedException {
    int sum = 0;
    for (IntWritable val : values) {
        sum = val.get();
    }
    context.write(key, new IntWritable(sum));
}

medium

A. The Iterable<IntWritable> should be a List<IntWritable>

B. The key parameter type is incorrect

C. The context.write method is missing

D. The sum variable is overwritten instead of accumulated

Solution

1. Step 1: Analyze the for loop logic

   The code assigns sum = val.get() inside the loop, overwriting sum each time.

2. Step 2: Identify correct accumulation

   Sum should accumulate values using sum += val.get() to add all values.

3. Final Answer:

   The sum variable is overwritten instead of accumulated -> Option D

4. Quick Check:

   Use += to accumulate sum, not = [OK]

Hint: Use += to add values, not = which overwrites [OK]

Common Mistakes:

• Using = instead of += in accumulation
• Confusing key type with error
• Thinking context.write is missing

5.

You want to count the total occurrences of each word in a large text dataset using Hadoop MapReduce. Which of the following best describes what the Reduce phase should do?

hard

A. Receive each word as key and list of counts as values, then sum counts to get total occurrences

B. Split the input text into words and assign counts

C. Sort all words alphabetically before counting

D. Write the raw text data to output without changes

Solution

1. Step 1: Understand MapReduce word count flow

   Map phase outputs (word, 1) pairs for each word occurrence.

2. Step 2: Role of Reduce phase in counting

   Reduce receives each word as key and list of counts (1s), sums them to get total count.

3. Final Answer:

   Receive each word as key and list of counts as values, then sum counts to get total occurrences -> Option A

4. Quick Check:

   Reduce sums counts per word [OK]

Hint: Reduce sums counts for each word key [OK]

Common Mistakes:

• Confusing Map and Reduce roles
• Thinking Reduce sorts data
• Assuming Reduce outputs raw data