Windowed operations in Kafka - Mini Project: Build & Apply

Windowed Operations with Kafka Streams

📖 Scenario: You are working with a stream of sales data coming from a Kafka topic. Each record contains a product name and the number of units sold. You want to calculate the total units sold for each product in 1-minute time windows.

🎯 Goal: Build a Kafka Streams application that reads sales data, groups it by product name, applies a 1-minute tumbling window, sums the units sold per product in each window, and outputs the results.

📋 What You'll Learn

Create a Kafka Streams builder and define the input stream from topic sales.

Create a 1-minute tumbling windowed aggregation grouped by product name.

Sum the units sold per product in each window.

Print the windowed aggregation results to the console.

💡 Why This Matters

🌍 Real World

Windowed operations are used in real-time analytics to summarize data over fixed time periods, such as counting sales per minute or monitoring sensor data.

💼 Career

Understanding windowed aggregations in Kafka Streams is essential for building scalable, real-time data processing applications in roles like data engineer or streaming developer.

1

Create the initial stream from the sales topic

Create a Kafka Streams builder called builder and define a stream called salesStream from the topic "sales" with key type String and value type Integer.

Kafka

// Create a StreamsBuilder and define salesStream from topic "sales"
// Your code here

Need a hint?

Use StreamsBuilder to create the builder and builder.stream("sales") to get the stream.

2

Define a 1-minute tumbling window for grouping

Create a variable called windowSize of type Duration and set it to 1 minute. Then create a TimeWindows object called tumblingWindow using TimeWindows.ofSizeWithNoGrace(windowSize).

Kafka

StreamsBuilder builder = new StreamsBuilder();
KStream salesStream = builder.stream("sales");
// Define a 1-minute Duration called windowSize
// Define a tumbling TimeWindows called tumblingWindow
// Your code here

Need a hint?

Use Duration.ofMinutes(1) for the window size and TimeWindows.ofSizeWithNoGrace(windowSize) for the tumbling window.

3

Group by product and sum units sold in each window

Group salesStream by key using groupByKey(), then apply a windowed aggregation using windowedBy(tumblingWindow) and sum the values using reduce(Integer::sum). Store the result in a variable called windowedSales of type KTable<Windowed<String>, Integer>.

Kafka

StreamsBuilder builder = new StreamsBuilder();
KStream salesStream = builder.stream("sales");
Duration windowSize = Duration.ofMinutes(1);
TimeWindows tumblingWindow = TimeWindows.ofSizeWithNoGrace(windowSize);
// Group salesStream by key, window by tumblingWindow, and sum values
// Your code here

Need a hint?

Use groupByKey() then windowedBy(tumblingWindow) and reduce(Integer::sum) to sum units sold per window.

4

Print the windowed aggregation results

Use windowedSales.toStream() and then foreach() to print each window's product name, window start time, and total units sold. Use System.out.println inside the lambda to print in the format: "Product: {key}, Window start: {windowStart}, Total units: {value}".

Kafka

StreamsBuilder builder = new StreamsBuilder();
KStream salesStream = builder.stream("sales");
Duration windowSize = Duration.ofMinutes(1);
TimeWindows tumblingWindow = TimeWindows.ofSizeWithNoGrace(windowSize);
KTable, Integer> windowedSales = salesStream
    .groupByKey()
    .windowedBy(tumblingWindow)
    .reduce(Integer::sum);
// Print each windowed aggregation result
// Your code here

Need a hint?

Use toStream() and foreach() to print each windowed result with System.out.println.