For example, if I want to store orders/booking information then I can use order creation date to index data into “order-yyyy” index. But is there any benefits in doing so? For log data using such indexes is a good idea because it is easier to remove old logs. But what about regular and also somewhat time-based data? I want to keep orders data forever.
So far I’ve come up with this: the only reason to do this is that each year this strategy will allow to automatically scale cluster and prepare it for new data, which will probably improve performance. But intuition suggests that manual fine grained resharding when it’s needed will be much better approach, and I better to stick with “orders-vX” naming schema (in case of mapping changes).
