chronon
chronon copied to clipboard
[CHIP] Fetch Stats NRT + Group By Stats
Problem Statement
Some of the more relevant stats such as score drift don't require a join to define. They can be defined a streaming aggregation of scores. It's desirable to have existing stats such as null counts and drift for log data to be able to be processed in microbatches into the KV Store with a reasonable cadence (say 5 minutes). This would allow to build near realtime monitoring that computes drift, null counts and other statistics on top of fetched data.
Requirements
- Fetch statistics from group bys.
Verification
- Unit tests
Approach
- TBD, could be an application of the Tiled aggregation or a separate streaming job.
User API (when required)
- Extension to existing endpoints in Fetcher API
Planning
TBD