Using replicas for long term storage in SAN

Search is usually quicker on hot nodes as they have more/faster resources to work with.

ES compressed data by default, that ratio depends entirely on the dataset.