Streaming Data Temporal Window Functions

 

https://learn.microsoft.com/en-us/azure/stream-analytics/stream-analytics-window-functions

 

Tumbling window

  • No overlap (one event belongs to one and only one window), no gaps.
  • Each window is the same size.
  • Windows can be empty.
  • Window start and finish is controlled by time.

 

Hopping window

  • Windows overlap, but there are no gaps.
  • Each window is the same size.
  • Windows can be empty.
  • Window start and finish is controlled by time.

Tumbling windows are just hopping windows with an overlap time of 0.

 

Sliding window

  • Windows overlap, and there can be gaps.
  • Each window is the same size.
  • Windows cannot be empty.
  • Window start and finish is controlled by events - logically considering all possible windows of a given length.

 

Session window

  • Windows do not overlap, and there can be gaps.
  • Window size is not constant (we define min and max).
  • Windows cannot be empty.
  • Window start is controlled by events. Window finish is controlled by events, within the min and max time.

 

Snapshot window

  • Windows do not overlap, and there are gaps.
  • Window size is essentially zero.
  • Windows cannot be empty.
  • Window start is controlled by events.