Kafka Streams 소개

Imagine data as a continuous flow, like a river. Stream processing is about analyzing and reacting to this data as it arrives, in real-time.

Unlike batch processing, which handles data in large, fixed groups (like a lake), stream processing works on individual data points or small windows of data as they are generated.

Kafka Streams is a client library for building powerful stream processing applications. It's an integral part of the Apache Kafka ecosystem.

It allows you to process data stored in Kafka topics, perform transformations, aggregations, and then write the results back to Kafka or external systems.

Kafka Streams simplifies building stream processing apps by:

• No Separate Cluster: It runs directly within your application, not on a dedicated processing cluster.
• Scalability: Inherits Kafka's distributed nature, scaling effortlessly with partitions.
• Fault Tolerance: Automatically handles failures and recovers state.
• Developer Friendly: Provides a high-level API for common operations.

In Kafka Streams, data flows as a stream, which is an unbounded, ordered, and re-playable sequence of data records.

Each record is a simple key-value pair. For example, a user ID could be the key, and a user action (like "logged in") could be the value.

An application built with Kafka Streams is often called a stream processor. It reads input streams from one or more Kafka topics, processes the data, and can produce output streams to other Kafka topics.

These applications can perform various operations, from simple filtering to complex stateful aggregations.

Let's look at the absolute minimum to get a Kafka Streams application running. You'll need a few configurations and a StreamsBuilder.

This example sets up the basic structure but doesn't perform any actual data processing yet. It just defines the "blueprint" of your stream application.

The Properties object holds crucial settings for your Kafka Streams application:

• APPLICATION_ID_CONFIG: Unique ID for your app within the Kafka cluster.
• BOOTSTRAP_SERVERS_CONFIG: List of Kafka brokers to connect to.
• DEFAULT_KEY_SERDE_CLASS_CONFIG: How keys are serialized/deserialized.
• DEFAULT_VALUE_SERDE_CLASS_CONFIG: How values are serialized/deserialized.

Serdes (Serializer/Deserializer) are vital. Kafka only understands bytes, so your application needs to convert Java objects (like Strings or Integers) into bytes before sending them to Kafka, and convert bytes back into objects when consuming.

Kafka Streams provides built-in Serdes for common types like String, Long, and Integer via Serdes.String(), Serdes.Long(), etc.

Kafka Streams is ideal for many real-time scenarios:

• Real-time Analytics: Monitoring dashboards, fraud detection.
• Data Transformation: Cleaning, enriching, and restructuring data streams.
• Event-Driven Microservices: Building reactive services that communicate via events.
• ETL Pipelines: Continuous extraction, transformation, and loading of data.

Great job! You've taken your first steps into Kafka Streams.

• We defined stream processing and introduced Kafka Streams as a powerful library.
• We explored its benefits like scalability and fault tolerance.
• You saw the fundamental concepts of streams, records, and essential configurations, including Serdes.

Next, we'll dive deeper into the core APIs: KStream and KTable, to start building actual data transformations!