One of the most commons reasons to use recurring big data analytics is to perform processing in near real-time. For example, a big data analytic configured to run every few minutes or hours which processes only the most recent features written and stored in a feature layer.
As another example, consider a real-time analytic configured to receive data from a feed that collects vehicle location updates every 10 seconds. This real-time analytic writes event data to a Feature Layer (new) output and calculates a date field (named something such as process_timestamp) using the Calculate Field tool with the time in which an event was processed using the Arcade Date() function.
It is a best practice to use the Calculate Field tool in a real-time analytic to write the datetime of processing to the feature layer that will be consumed by the big data analytic for near real-time analysis. The reason being, some data sources consumed by feeds have an inherent delay in providing data or polling that could cause features to be missed by the timestamp field queries.
To complement this real-time analytic, a scheduled recurring big data analytic can be configured that uses the output of the real-time analytic as its data source. In this recurring big data analytic, a Feature Layer source would be configured to ingest the feature layer output created by the real-time analytic. When configuring the Feature Layer source, in the Timestamp Field step, a date field can be selected as the timestamp field. Select the datetime field created by the Calculate Field tool in the real-time analytic, in this case the field is process_timestamp.
A timestamp field is utilized by the Feature Layer source to retrieve only the latest features from the feature layer on each run. If a timestamp field is specified, the first time ArcGIS Velocity polls the feature layer it will load all features with a timestamp field datetime less than the first scheduled run time that also meet the criteria of the WHERE clause. With each subsequent run, features with a timestamp field value between the last scheduled run time and the current scheduled run time that also meet the criteria of the WHERE clause will be loaded.
The big data analytic would be configured to run at the desired repeat interval such as every 5 minutes. Using the timestamp field as outlined above, only the most recent features not yet processed would be analyzed by the big data analytic during subsequent runs.