We’ve not tested CE at that scale, so it’s difficult for us to characterize without more information on what the cluster would be doing. As I’m sure you know, Enterprise Edition has a bunch of performance optimizations that don’t exist in CE like Multidimensional Scaling (the ability to allocate and isolate how much resource each service consumes on a node), the Plasma storage engine for Indexing, the query engine optimizations, etc… You would likely find, that your cost/benefit would indicate that it would be better to use EE, which could get you to 1700 RPS on significantly fewer instances, and more than pay for itself. You could do a quick, cheap test of this using the Couchbase Cloud trial, too. We do know that EE clusters have gone into hundreds of nodes.