Data can be categorized as bounded or unbounded. Bounded data is data at rest and is usually processed through a batch process. Streaming is basically data processing on unbounded data. Let's look into an example. Let's assume that we are analyzing fraudulent transactions at a bank. If we want to look for fraud transactions 7 days ago, we have to look at the data at rest; this is an example of a batch process.

n the other hand, if we want to detect fraud in real-time, that is an example of streaming. Hence, streaming data algorithms are those algorithms that deal with processing data streams. The fundamental idea is to divide the input data stream into batches, which are then processed by the processing node. Streaming algorithms need to be fault-tolerant and should be able to handle the incoming velocity of data. As the demand for real-time trend analysis is increasing, the demand for stream processing is also increasing these days. Note that, for streaming to work, data has to be processed fast and while designing algorithms, this needs to be always kept in mind.