Skip to content

Latest commit

 

History

History
25 lines (13 loc) · 3.03 KB

Fundamentals.md

File metadata and controls

25 lines (13 loc) · 3.03 KB

Fundamentals of Streaming

Streaming in software development refers to the real-time processing of data, in which data is continuously generated, transmitted, and consumed by various applications and services. This process enables businesses to process large volumes of data in real-time, providing valuable insights and improving decision-making.

Key concepts in streaming:

  1. Data streams: Data streams are continuous sequences of data records. They can be generated from various sources, such as user interactions, IoT devices, logs, and sensors. Data streams are usually unbounded, meaning they have no fixed beginning or end.

  2. Stream processing: Stream processing is the act of processing data as it arrives, often in real-time. It involves reading data records from a data stream, processing them, and optionally writing the results to another data stream or storage system. Stream processing can handle tasks such as filtering, aggregating, joining, and windowing data.

  3. Real-time vs. batch processing: Traditional batch processing processes large volumes of data at periodic intervals, while real-time stream processing processes data as it is generated. Real-time processing is beneficial for time-sensitive applications and for reducing latency in data processing.

  4. Producers and consumers: Producers generate data and send it to a data stream, while consumers read and process the data. Both producers and consumers can be distributed across multiple systems for scalability and fault tolerance.

  5. Stream processing frameworks: Various frameworks and tools exist for stream processing, including Apache Kafka, AWS Kinesis, and Azure Event Hubs. These tools provide infrastructure for creating, managing, and processing data streams.

  6. Windowing: Windowing is the process of grouping data records in a stream based on time or other criteria, such as event count. This allows for processing and analysis of data within defined windows, making it easier to detect patterns and trends.

  7. Stateful vs. stateless processing: Stateless processing does not maintain any information about previous data records, while stateful processing keeps track of the state across multiple records. Stateful processing allows for more complex operations, such as computing rolling averages or detecting event sequences.

  8. Fault tolerance and scalability: Streaming systems must be able to recover from failures and scale to handle increasing data loads. This can be achieved through techniques such as replication, partitioning, and checkpointing.

  9. Backpressure: Backpressure is a mechanism used to prevent data overload in streaming systems. When consumers cannot keep up with the rate of incoming data, backpressure signals producers to slow down or buffer data until the consumers can catch up.

  10. Exactly-once processing: Ensuring that each record in a data stream is processed exactly once is important for maintaining data consistency and correctness. This can be challenging in distributed systems, but some streaming platforms offer built-in support for exactly-once processing.