Understanding data streams

A data stream is the continuous flow of any type of data using any medium. Out of 4 Vs of big data, two are velocity and variety. A data stream refers to both velocity and variety of data. Data stream is real-time data coming from sources such as social media sites or different monitoring sensors installed in manufacturing units or vehicles. Another example of streaming data processing is IOT, that is the Internet Of Things, where data is coming from different components though the internet.

There are two different kinds of streaming data: bounded and unbounded streams, as shown in the following images. Bounded streams have a defined start and a defined end of the data stream. Data processing stops once the end of the stream is reached. Generally, this is called batch processing. An unbounded stream does not have an end and data processing starts from the beginning. This is called real-time processing, which keeps the states of events in memory for processing. It is very difficult to manage and implement use cases for unbounded data streams, but tools are available which give you the chance to play with them, including Apache Storm, Apache Flink, Amazon Kinesis, Samaza, and so on.

We will discuss data processing in the following chapters. Here, you will read about data ingestion tools which feed in to data processing engines. Data ingestion can be from live running systems generating log files or come directly from terminals or ports.