Skip to main content

Streaming Analytics and Data Pipelines

Updated Sep 21, 2020 ·

Overview​

Streaming analytics processes data continuously instead of waiting for batch jobs.

  • Handles events as they happen
  • Reduces analysis delay
  • Supports time-sensitive decisions

This is useful for data sources like sensors, clickstreams, app activity, and financial feeds.

Batch Versus Streaming​

Batch processing groups data and processes it later, while streaming processes records as they arrive.

  • Batch is good for scheduled workloads and large periodic jobs
  • Streaming is good for real-time monitoring and fast reactions

Streaming analytics is the better fit when the value of the data drops quickly over time.

Pub/Sub​

Pub/Sub is Google Cloud’s publish and subscribe messaging service.

  • Receives messages from many producers
  • Distributes data to one or more subscribers
  • Works well for asynchronous event streams

Pub/Sub is often the first step in a real-time pipeline because it collects incoming events reliably.

Dataflow​

Dataflow processes both streaming and batch data pipelines.

  • Uses Apache Beam as the pipeline model
  • Handles ETL-style processing
  • Scales automatically with pipeline demand

Dataflow is a managed service that reduces the infrastructure work required to build and operate pipelines.

Typical Pipeline​

A common streaming flow looks like this:

  • Applications or devices send events to Pub/Sub
  • Dataflow transforms and enriches the events
  • BigQuery stores the processed data for analysis
  • Looker visualizes the results for business users

This flow supports real-time insights without forcing teams to wait for a manual batch process.

Common Use Cases​

Streaming analytics is useful when events need immediate action.

  • Monitor sensors and operational systems
  • Detect unusual account activity
  • Track clickstreams and user behavior
  • Enrich events for reporting and dashboards

Note: Streaming is not only about speed. It is also about keeping data fresh enough to drive the decision that depends on it.