Hi again! I'm sorry for being absent for so long. Maybe the reason here that you data is collocated somehow using id of a city. Replicated cache suits well if you need to join to some static dictionary and it's impossible or difficult to distribute other data in manner corresponding to that dictionary. My assumption here is that Ignite joins only subsets (about 1/2) of city cache in partitioned case. But when you have replicated one the Ignite should join entire dataset. What do you think does it make sense here?
-- Sent from: http://apache-ignite-users.70518.x6.nabble.com/
