My understanding is that parallel execution (on multiple nodes) only happens when in partitioned mode.
>From a performance point of view it would be nice if queries running on fully replicated records can be, optionally, executed by all nodes to improve performance. Basically the query could be split up the same way as in partitioned mode, i.e. running on each node only for the records it is the master of. The only thing different is there is no need for affinity keys as all data is available on each node. Or is this too simplistic a thought? Paulus de B. -- Sent from: http://apache-ignite-users.70518.x6.nabble.com/