One of the attractions of the current "big data" fad is scalability - you can throw ten or a hundred servers at the problem of persisting data and then (mostly) selecting from it.
Seems to me that we could do that with SQL Server - have a single (cluster) host where data is inserted and updated, and have multiple read-only copies to support scalability to thousands of simultaneous select/SP transactions.
But I haven't seen it done. This may or may not involve sharding, where a copy contains only a part of the data, and may or may not - but mostly not - involve letting the clients do any data modification.
"Scaling out" and sharding were somewhat more popular ten years and more ago, before SQL Server and Wintel got so good at multiple processors per server. Anybody doing any of that now?
Outside of log shipping to DR and replication to a report server, I haven't seen any recently.
Shipping data from an OLTP system to a data warehouse or data mart via an operational data store does not count, that's changing the shape of the data as it moves and not using the destinations homogenously.
Thanks,
Josh