Example: Elasticsearch has NRT (Near Real Time) search capabilities. Typical indexing (or optimmizing existing fragmented index) of large datasets will take hours; if we need to index latest tweets from Twitter in real-time, for past few hours only, we only need milliseconds, so that it is called NRT; and it supports near-real-time updates of the index and doesn’t need any “optimization”.
Now, what about Near-Real-Time Search for past week, month, year? Index optimization will take hours, and we have realtime data updates, it won’t be optimized and we can’t use NRT.
Solution: implement five services and aggregate results: