With ducklake this scales well to multi-terabyte data sets. The big benefit of this server protocol is sharing a high memory server and taking advantage of a shared cache for recent data.