The historical data view (last few days) architecture is not scalable by any means. It currently reads a bunch of individual blobs from storage, creates a historical time window, and caches it. Would likely by much better served by a Function that updates a fixed data set, either in blob (as avro or json), or in something like Cosmos, keeping the historical read lightweight and simple.
Need an approach for this "cold path" data that keeps it updated and lightweight, and removes the burdon on computing this on every client implementation.
The historical data view (last few days) architecture is not scalable by any means. It currently reads a bunch of individual blobs from storage, creates a historical time window, and caches it. Would likely by much better served by a Function that updates a fixed data set, either in blob (as avro or json), or in something like Cosmos, keeping the historical read lightweight and simple.
Need an approach for this "cold path" data that keeps it updated and lightweight, and removes the burdon on computing this on every client implementation.