At some point, to handle a big chunk of data, you need to think about data-warehouse, move the resent and "hot" event to a more "on memory" db and let the old, less use data at the end of the pipeline. Memory DB => Normal DB => Store DB and take particular attention to the communication between them.