Windowing

Continuously streaming data makes real-time computations and insights possible, overcoming the latency inherent in batch data processing systems. However, insights requiring aggregations of data, even very recent trending (for example, in the past 10 seconds), need to break the data stream into bounded groups of events. Time is a fundamental concept of streaming data systems and the natural construct to use when defining event boundaries for computing aggregations.

The following screenshot shows an event stream with defined time windows overlayed and sample computations produced. Note that the time windowing is fundamental to computing aggregates, like a count of events:

Stream Analytics uses windows of time to group events and supports window types that enable a variety of common event grouping patterns. In this section, we will examine the tumbling window, hopping window, and sliding window types. Stream Analytics windows are always used in the GROUP BY query clause. Queries will...