In terms of improvements for Denodo, regarding performance, in cases where there are multiple virtualizations—such as reading from one Denodo view that is virtualized, and from that view there's also virtualization, and another team is reading from that view—if multiple virtualizations happen with no caching in between, it becomes slow. This occurs because it is cascading; whenever at the top level someone is reading data, that request is getting cascaded to the nth level, causing issues, especially in cases such as Power BI reports. We need to consider implementing some persistent layer in between. The scaling process should improve because many things are getting automated. The scale-out part needs to be automated, though I am uncertain whether Denodo has already implemented that feature.
